Design of a Humanoid Biped for Walking Research ...Design of a Humanoid Biped for Walking Research...
Transcript of Design of a Humanoid Biped for Walking Research ...Design of a Humanoid Biped for Walking Research...
Design of a Humanoid Biped for Walking Research
by
Daniel Joseph Paluska
Submitted to the Department of Mechanical Engineeringin partial fulfillment of the requirements for the degree of
Master of Science
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
September 2000
c© Massachusetts Institute of Technology 2000
Signature of Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Department of Mechanical Engineering
August 31, 2000
Certified by. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Gill A. Pratt
Assistant Professor of Electrical Engineering and Computer Science, MITThesis Supervisor
Certified by. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Ernesto Blanco
Adjunct Professor of Mechanical Engineering, MITThesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Ain Sonin
Chairman, Departmental Committee on Graduate Students
Design of a Humanoid Biped for Walking Researchby
Daniel Joseph Paluska
Submitted to the Department of Mechanical Engineeringon August 31, 2000, in partial fulfillment of the
requirements for the degree ofMaster of Science
Abstract
This thesis presents the design of the robot M2. The primary motivation behind the work is tocreate a platform for research into bipedal walking. Bipedal robots have been built in the past, butmost of them relied heavily on lessons from robotic arm research. There is evidence that suggestsactuation and design for legged locomotion should deviate from the past standard of robotic armdesign. Specifically, there should be an emphasis on low impedance actuation, shock tolerance,passive control mechanisms, and weight reduction. Human data which supports these points ispresented.
The robot M2 was designed with these things in mind. M2 is a humanoid bipedal robot withtwelve active degrees of freedom. It is essentially a torso and two legs and with linear dimensions ofa 50th percentile US male. It weighs approximately 60 lbs(25kg). All the active degrees of freedomare powered using Series Elastic Actuators, which provide force control and shock tolerance. Forsimplicity, degrees of freedom above the hip are absent.
Thesis Supervisor: Gill A. PrattTitle: Assistant Professor of Electrical Engineering and Computer Science, MIT
Thesis Supervisor: Ernesto BlancoTitle: Adjunct Professor of Mechanical Engineering, MIT
3
Acknowledgments
The development of the robot has been a three person project from the beginning. Jerry Prattand David Robinson worked on the robot with me. Jerry performed preliminary simulations todetermine feasibility of mass distributions and power requirements. David Robinson is responsiblefor the actuator design. I also had lots of help from Ben Krupp and Chris Morse who each had thereown robot to build in the process. Pete Dilworth contributed plenty of tips from his own bipedalrobot building experience. Allen Parseghian has been working hard with me trying to get the robotto actually walk. John Hurst has helped with debugging some actuator problems and designing abattery rig. Thanks to Mike Wessler and Andreas Hofmann for creation of and help with softwaretools. Thanks to Chris Barnhart for developing the DSP platform and handling my ignorance of allthings DSP.
During the construction, I had help from many people. We had several soldering parties in thelab which involved people sitting around the main lab table soldering. Jerry, Ben, Chris, Mike,Gaddy, Pete, and others participated in these. In addition, Jess helped me with cable making andwas an immense help with my supposed native language.
Professor Pratt has been a wonderful advisor. I am constantly amazed at how he has alwaysplaced his students first. I have loved working in the lab and getting to play around!
Special thanks goes to my friends and family who have made the time outside the lab great!You’ve kept me sane during the worst of thesis despair.
This research was supportd by DARPA under contract number N39998-00-C-0656.
Contents
1 Introduction 91.1 Goals of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.2 Summary of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Human Characteristics 112.1 The Human Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Weight Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1.2 Link Lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Human Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.1 Energetics of Human Walking . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.2 Human Kinetic and Kinematic Data . . . . . . . . . . . . . . . . . . . . . . . 15
3 Robots and Actuators 163.1 Robotic Arms vs. Robotic Walkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163.2 Passive Dynamic Walkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163.3 Powered Bipedal Walkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.4 Series Elastic Actuators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4 Robot Mechanical Design 194.1 General Robot Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.2 The Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2.1 Overall Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194.2.2 Actuators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224.2.3 Ankle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224.2.4 Foot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224.2.5 Hip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244.2.6 Knee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Electronics and Control Systems 275.1 Electronics Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.1.1 48 Volt bus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275.1.2 Robot Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2 Custom Circuit Boards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.2.1 The Signal Conditioning Boards . . . . . . . . . . . . . . . . . . . . . . . . . 295.2.2 The Analog Force Control Boards . . . . . . . . . . . . . . . . . . . . . . . . 295.2.3 The Analog Breakout Board . . . . . . . . . . . . . . . . . . . . . . . . . . . 315.2.4 The Power Distribution Boards . . . . . . . . . . . . . . . . . . . . . . . . . . 315.2.5 The DSP and Analog I/O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.3 Control Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4
CONTENTS 5
A Joint Math 34A.1 Knee Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34A.2 Ankle Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36A.3 Hip Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
A.3.1 Hip Pitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37A.3.2 Hip Roll and Yaw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
B Electrical Schematics 39
C Suppliers and Costs 50C.1 Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
List of Figures
2-1 The X velocities of the various parts of the body during a normal walking cycle.Adapted from data in Winter(20). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2-2 The frontal plane dimensions of a 50th percentile U.S. male. Adapted from data inDreyfus(2). All dimensions are in inches. . . . . . . . . . . . . . . . . . . . . . . . . . 13
2-3 Several strides of the compass gait shown with corresponding potential and kineticenergy curves. From Krupp(8). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3-1 Some previous powered bipedal walking robots. . . . . . . . . . . . . . . . . . . . . . 173-2 An annotated rendering of the Series Elastic Actuator which is used in the robot M2.
Modeled and rendered by David Robinson. . . . . . . . . . . . . . . . . . . . . . . . 18
4-1 A photo of the completed robot and a joint schematic view of the robot. The schematicshows active degrees of freedom only. The optional passive toe joint is not shown. . . 20
4-2 The dimensions of the biped in the frontal plane. The body and links are approxi-mately axially symmetric about their longitudinal axis. Dimensions are in inches. . 21
4-3 A schematic of the ankle joint actuation scheme. The axes shown are fixed to thecenter of the universal joint. The points A,E, and O are referenced in Appendix A. . 23
4-4 The dimensions of the biped foot with passive toe joint. The X’s represent the locationof the load cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4-5 A schematic of the biped hip. The pitch actuator is not shown. It lies along the axisof the thigh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4-6 A photo of the biped thigh illustrating the hip pitch cable drive system. . . . . . . . 244-7 The ranges of motion on the robot ankle joint. The ankle roll is symmetric(+- 20 deg). 254-8 The range of motion of the robot knee. The knee has 80 degrees of motion. . . . . . 26
5-1 An overview of the biped electronics. Thick lines indicate power transfer and arrow-heads indicate information flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5-2 A photo of M2 with annotations of electrical systems. . . . . . . . . . . . . . . . . . 295-3 A photo of the biped strain gauge signal conditioning board. This board handles two
strain gauges. There are a total of four of these boards on the robot. . . . . . . . . 305-4 A photo of the analog force control and joint pot buffer board. There are six identical
boards on the robot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
A-1 The robot knee joint with superimposed lines and points. Points O, A and M all referto pin joints. O is the knee joint. M is where the actuator is mounted to the thighand A is where the actuator is attached to the shin. . . . . . . . . . . . . . . . . . . 35
A-2 A line drawing of the knee for the calculation of knee actuator desired force. Theactuator is the line segment �MA. This drawing also pertains to the geometry of thehip and ankle joints but the hip and ankle actuators have motions out of the planewhereas the knee actuator is always in the plane perpendicular to the knee axis. . . 35
6
LIST OF FIGURES 7
A-3 A simple view of the actuation of the robot ankle. This is a tranverse plane slice ofthe robot ankle. Point O is the intersection of the X and Y axes as well as the centerof the ankle universal joint. Points A1 and A2 represent the attachment points of thetwo actuators respectively. Forces in the actuators create torques about both the Xand Y axes. Joint rotations change the relative lengths of the X and Y projections(rx
and ry) of the moment arms, therefore affecting the torque to force transformation. . 36A-4 The robot hip joint schematic. Point O refers to the hip universal joint. The Z(hip
yaw) and Y(hip roll) axes intersect at point O. Point Mr refers to the universal jointon the body where the roll actuator is mounted. Point Ar refers to the ball joint onthe thigh where the roll actuator is attached. Point My refers to the pin joint wherethe yaw actuator is attached to body. Point Ay is the pin joint attachment of the yawactuator to the yaw universal block. It is hidden in this schematic. . . . . . . . . . 37
B-1 Page 1 of 6 of the biped force board. . . . . . . . . . . . . . . . . . . . . . . . . . . . 40B-2 Page 2 of 6 the biped force board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41B-3 Page 3 of 6 of the biped force board. . . . . . . . . . . . . . . . . . . . . . . . . . . . 42B-4 Page 4 of 6 the biped force board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43B-5 Page 5 of 6 of the biped force board. . . . . . . . . . . . . . . . . . . . . . . . . . . . 44B-6 Page 6 of 6 the biped force board. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45B-7 Page 1 of 2 of the strain gauge conditioning board. . . . . . . . . . . . . . . . . . . . 46B-8 Page 2 of 2 of the strain gauge conditioning board. . . . . . . . . . . . . . . . . . . . 47B-9 Page 1 of 1 of the analog I/O breakout board. This board is designed to interface
with the ANA070 and ANA064 from Digital Designs and Systems. . . . . . . . . . . 48B-10 Page 1 of 1 of the M2 power distribution board. This board has two different popu-
lation options. One to power the DSP and one to power the Intersense IS-300 tracker. 49
List of Tables
2.1 Human Mass Distribution. From Dempster(3). . . . . . . . . . . . . . . . . . . . . . 112.2 Human walking parameters from normalized data contained in Human Walking(18).
Forces and power calculated for a for 1.83 m(6’2”), 80 kg(178lb) person. Table displaysnon-concurrent maximum values which occur during an average walking cycle. . . . 15
4.1 Robot Joint Specifications. Torque and rad/s numbers are given for maximum mo-ment arm. Power, torque, and velocity are symmetric due to the actuator. *Theankle roll and pitch are not independent. Their maximum values can not be appliedsimultaneously. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Approximate distribution of mass in humans, the biped M2, and McGeer’s kneed pas-sive dynamic walker(PDW). Human data adapted from Dempster and Gaughran (3).Robot weight distribution is driven primarily by actuator locations. Each actuator isapproximately 1.2kg(2.5 lbs). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.1 Circuit Boards on the Robot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275.2 Sensors on the Robot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
C.1 Basic robot budget. See Robinson et al.((17)) for more detail on the actuator. Thisbudget does not include prototyping or development costs. Part quantities and indi-vidual costs are not necessarily meant to imply identical parts but rather to give anaverage price for all units of a certain type. . . . . . . . . . . . . . . . . . . . . . . . 54
8
Chapter 1
Introduction
1.1 Goals of Thesis
The goal of this research project was to design a three dimensional, free standing bipedal robot. Themain goal of the biped, M2, is to be a testbed for control ideas and walking research. Originallythe robot was not specifically ’humanoid,’ but after a bit of research we determined it would be thebest course of action for several reasons.
• There is a vast amount of data regarding human form and motion.
• The robot can be considered a human replacement for hazardous situations.
• The robot will be large enough to avoid problems of miniaturization.
• Ideas regarding human control of locomotion can be tested.
This thesis concentrates on documenting the mechanical portion of the project. The design ofthe robot was guided by several specific goals which include walking 1 m/s, climbing normal stairs,looking biological, turning dynamically, a three year life-span, and ten hours working time betweenmechanical or electrical failures.
The following points dominated the design decisions of the robot.
1. Series Elastic Actuators Series Elastic Actuators(12) are used for all of the active degreesof freedom. These actuators provide force control as well as shock tolerance. Evidence sug-gests both factors are necessary for the task of biologically similar walking. The low outputimpedance of the actuators allows us to take advantage of the robot’s natural dynamics. Alljoints employ the same actuator design to minimize complexity and facilitate repairs.
2. Human Proportions The use of human proportions allows for easy comparison with biome-chanics data. Human sizing also allows for use of large, standard components which are easyto see and debug. Also, by using human proportions, the research stays focused on walkingand not on miniaturization.
3. Lightweight The robot frame is carbon fiber and most other components are plastic or alu-minum. The necessary actuator forces are kept low. A less massive robot is more manageablein a research environment. It is easier to handle and less likely to damage itself or harmresearchers.
4. Mechanical Control Mechanisms Each joint (most importantly the knee) has adjustablestops with rubber pads. The foot of the robot is equipped with a passive toe joint. Thisjoint has an adjustable range as well as a return spring. The limit stops are essentially highfrequency non-linear PD loops which are difficult to implement in digital control even with theuse of sophisticated electronics and sensors. The low impedance actuators allow for uncertain
9
CHAPTER 1. INTRODUCTION 10
contact, an inevitable occurence during walking on rough terrain. These mechanical featureseliminate the need to operate any digital high frequency control loops on the robot.
The robot is a three dimensional continuation of the work that began with the planar robot SpringTurkey(13) and continued with the planar robot Spring Flamingo(15). The previously enumeratedpoints are the key areas where M2 differs from some other current three dimensional humanoidwalkers(6, 21) which rely more on high impedance actuation and trajectory following control schemes.
1.2 Summary of Thesis
This thesis presents the design of the humanoid robot M2.
• Chapter 2 describes the human mechanisms and the forces, torques, velocities, etc. which areexhibited during a normal walking cycle. The human is a good starting point for the designof the robot. Human locomotion is well documented and there is a wealth of information onthe human walking motion.
• Chapter 3 describes previous robots and actuators. It presents some of the differences betweenrobotic arm and walking robot technology. There is a brief introduction to Series ElasticActuators.
• Chapter 4 describes the mechanical specifics of the robot.
• Chapter 5 details the electronic, sensor, computer, and control systems on the robot.
• Chapter 6 gives some conclusions and some advice for future research.
• Appendix A describes the transformations from joint torques to actuator linear forces.
• Appendix B contains all the schematics for the custom circuit boards on the robot.
• Appendix C has a list of all of the suppliers that were used.
Chapter 2
Human Characteristics
There is a large database of knowledge regarding both the human form and human motion. Webelieve that mechanical characteristics found in humans aid in the control of walking and we shouldtry to consider which of these characteristics can be embodied with current mechanical and robotictechnologies.
2.1 The Human Form
2.1.1 Weight Distribution
An approximation of the human weight distribution is shown in Table 2.1.From this we can see that nearly 70% of the mass in humans is located at the waist and above.
Figure 2-1 shows the linear velocities(in the walking direction, X) of different body areas. It canbe seens that the body has the least fluctuation in forward velocity and the foot has the mostfluctuation.
Since kinetic energy is a direct function of velocity, the lower extremities have the greatestfluctuations in kinetic energy. It is clear that more mass at the feet means more energy change.If this energy is all lost it would be very costly. In efficient walking, the mass is able to resonatewith gravitational or elastic potential energy. An example of this can be seen in the compass gaitdescribed in the next section.
2.1.2 Link Lengths
The dimensions of a 50th percentile male are shown in Figure 2-2. The center of gravity is justabove the hip at approximately 38”.
Table 2.1: Human Mass Distribution. From Dempster(3).
Body Area Percentage per Part No. of Instances Total Mass PercentageHead, neck, Shoulders, Thorax 31% 1 31%
Arm 5% 2 10%Ab/Pelvic 27% 1 27%
Thigh 10% 2 20%Shin and foot 6% 2 12%
11
CHAPTER 2. HUMAN CHARACTERISTICS 12
-1
0
1
2
3
4
5
1 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Frame Number
Vel
oci
ty(m
/sec
)
ribcage
hip
shin top
knee
ankle
toe
Figure 2-1: The X velocities of the various parts of the body during a normal walking cycle. Adaptedfrom data in Winter(20).
CHAPTER 2. HUMAN CHARACTERISTICS 13
0
3.20
19.80
36.50
60.20
(Hip Spacing)7.00
3.90
16.70
16.60
6.10
69.10
14.20
56.70
Figure 2-2: The frontal plane dimensions of a 50th percentile U.S. male. Adapted from data inDreyfus(2). All dimensions are in inches.
CHAPTER 2. HUMAN CHARACTERISTICS 14
KE
PE
Figure 2-3: Several strides of the compass gait shown with corresponding potential and kineticenergy curves. From Krupp(8).
2.2 Human Motion
2.2.1 Energetics of Human Walking
Data in Rose et al.(18) shows us that a 180lb person walking at a normal pace consumes about 320Watts of power. It is estimated that roughly one quarter of that is mechanical power. This meansthat at a comfortable pace, a 180lb human can walk with only 80 Watts of mechanical power. Thiscan be a somewhat depressing fact considering one of the electric motors used on the robot M2 canoutput 90 Watts continuously - and there are 12 motors on the robot! Hopefully research on M2will allow future robots to walk as mechanically efficient as a human.
A simplified human gait known as a compass gait can give us insight into the energetics ofwalking. The compass gait model consists of two rigid, massless legs attached at the hip with a pinjoint. A point mass is modeled at the hip joint. With the only degree of freedom located at the hip,the body is forced to follow an arc defined by the length of the leg. The compass gait can be seenin Figure 2-3.
The compass biped has instantaneous, discretely changing dynamics. It simply changes fromone inverted pendulum to another when the swing leg touches the ground. Calculating the externalwork on the model, namely the potential and kinetic energy of the system
PE = MgLcosθ (2.1)KE = MgL(1 − cosθ) (2.2)
By graphing the equations 2.1 and 2.2 in Figure 2-3, the cyclical nature of potential energy andkinetic energy can be seen. Note that the kinetic energy is maximum when the potential energy isminimum and vice versa. Similar cyclical transition of potential and kinetic energy was recorded
CHAPTER 2. HUMAN CHARACTERISTICS 15
deg rad/s Nm Whip(pitch) 30,-18 3.6,-2.0 -111.0 56hip(roll) 8,-7 1.6,-1.0 -63.5 +-28hip(yaw) 5,-15 4.0,-3.0 8.0 -16
knee(pitch) 68,8 5.8,-7.8 -71.4 -79.5ankle(pitch) 10,-15 3.0,-4.2 -63.5 280ankle(roll) na NA 40.0 -16
Table 2.2: Human walking parameters from normalized data contained in Human Walking(18).Forces and power calculated for a for 1.83 m(6’2”), 80 kg(178lb) person. Table displays non-concurrent maximum values which occur during an average walking cycle.
in human subjects by McMahon (11) using force plate data to calculate the changes in mechanicalenergy of the body’s center of mass.
The compass gait is similar to the behavior found in the passive dynamic walkers described in thenext section. There is also evidence that humans use some passive dynamics during the gait cycle. Atthe beginning of swing phase, the leg receives a tiny impulse of energy and then the natural pendulummotion carries it forward to extension. The leg receives another tiny impulse of energy to bring itto a sudden halt before touchdown. Electromyographic data record by McMahon (11) shows littleelectrical activity in the leg muscles of humans during the swing phase at normal walking speeds.This suggests that the leg is swinging freely during this period. In addition, electromyographicrecords show significant electrical activity in leg muscles during stance.
2.2.2 Human Kinetic and Kinematic Data
There is a wealth of data from recordings of human motion. These recordings have been transferredinto data regarding joint motions and joint torques during the walking cycle. In addition, theauthor relied on a rather low-tech method for extracting velocity data from the graphical positiondata. Power, torque, position, and velocity data for an average human walking cycle are shown inTable 2.2.
Chapter 3
Robots and Actuators
3.1 Robotic Arms vs. Robotic Walkers
Many high performance robotic arms and hands have been developed for use in factories, space, andresearch(19). It might seem to an outside observer that these technologies could be exploited foruse in a legged robot. Most of the time this is not the case. There are several reasons why.
• Fixed vs. Floating Reference Robotic arms are generally fixed to an inertial referenceframe(factory) or a body whose mass is large enough that it can be considered fixed(spacecraft).A walking robot is not fixed to any reference frame and has a limited set of torques which itcan apply due to its limited contact with the world.
• Onboard vs. Offboard Robot arms can often place their heavy motors at their fixed end.Then the motors are only responsible for moving the frame of the arm and not themselves.Because a walking robot must carry all its components, the motors support themselves as wellas the structure of the robot. Carrying power is also an issue for walking robots although mostare tethered due to battery limitations.
• Environmental Awareness Robot arms are not usually expected to perform in unknownsituations. They generally are designed for specific working conditions and their ability tohandle unexpected disturbances is limited. Ideally, walking robots are supposed to handlerough and unknown terrain.
• Success Metrics Robotic arms are often judged on their ability to position their end effectorsprecisely. Robotic walkers are usually not judged on their ability to position precisely butrather on their ability to get from Point A to Point B without falling down.
• Impacts Most robot arms are not designed to handle impacts. Walking, however, has animpact at every touchdown.
3.2 Passive Dynamic Walkers
Much research has been done on completely passive walking machines (5, 1, 4). Passive dynamicwalkers use Earth’s gravity as a power supply. They rely on special geometry and mechanical mech-anisms as control systems to achieve gaits very similar to the compass gait depicted in Figure 2-3.Though they are not robots in the traditional sense, they can give insight into walking machines.These ’robots’ exploit passive dynamic elements and mechanical mechanisms to achieve highly effi-cient gaits. In the limit, a passive dynamic walker relies entirely on passive dynamic motions andthus requires no external power source other than a small incline. Several passive legged ’robots’such as these have been constructed and succesfully demonstrated (5, 1, 4, 9).
16
CHAPTER 3. ROBOTS AND ACTUATORS 17
Figure 3-1: Some previous powered bipedal walking robots. From top left to bottom right areWL-10RV1 from Waseda, P2 from Honda, Toddler from UNH, the Moscow State University Biped,SD-2 from Clemson and Ohio State, Biper from University of Tokyo, Meltran II from MechanicalEngineering Lab in Tsukuba, and Timmy from Harvard. Figures compiled by J. Pratt.
3.3 Powered Bipedal Walkers
Many bipedal walking robots have been built over the years. Several of these robots are shown inFigure 3-1. These biped walking robots fall into two broad categories: those which predominantlyuse pre-recorded trajectory playback and those which predominantly use an algorithmic controller.All the robots employ a high impedance actuation scheme and none take advantage of naturaldynamics.
Passive dynamics were successfully implemented in actuated robots as well. J. Pratt (16) de-veloped algorithms that used the natural dynamics of the swing leg to efficiently control a planarbipedal robot. More recently, J. Pratt has developed three-dimensional simulations of M2 usingpassive elements such as a knee cap, a compliant ankle, and a passive swing leg to achieve naturallooking and efficient walking.
3.4 Series Elastic Actuators
Series Elastic Actuators(SEA’s)(14) are actuators which have an elastic element in series with themotor and gear train. A sensor measures the displacement of the elastic element and force is impliedby Hooke’s Law, F = kx. In short, SEA’s provide force control, shock tolerance, and low impedanceactuation. A more detailed account of their benefits and limits can be found in Robinson et al.(17)
A rendering of the SEA used on M2 can be seen in Figure 3-2. The brushless DC motor ismodified to have a ballscrew as its rotor. The output of the ballscrew is attached to an aluminum
CHAPTER 3. ROBOTS AND ACTUATORS 18
springs
brushless DC motor
linearpotentiometer
carbon fiber tubes
ballscrew
machined aluminum pieces
linear bushings
output shaft
Figure 3-2: An annotated rendering of the Series Elastic Actuator which is used in the robot M2.Modeled and rendered by David Robinson.
piece which is sandwiched between a set of linear compression springs. The other ends of thesprings are attached to the output of the actuator. A linear potentiometer measures the ballscrewdisplacement with respect to the output shaft, thereby measuring the spring compression and givingan indication of the force output.
On the specific SEA designed for M2(17), a simple analog PD controller is implemented to controlthe spring deflection. When the output of the actuator is clamped, a force control bandwidth of30hz is observed. The actuator is capable of outputting 300 lbs and capable of resolving forces onthe order of 1.5 lbs.
Chapter 4
Robot Mechanical Design
This chapter presents the design specifics of the bipedal robot M2.
4.1 General Robot Architecture
The robot stands approximately 5 feet high and weighs 60 pounds. The robot has 12 active degreesof freedom. Table 4.1 lists the specifications for the joints of the robot. Power at each joint is thesame due to the single actuator design employed for all the joints. The power to the individualdegrees of freedom of the ankle joint is limited by the power of the other.
The robot joint ranges of motion were chosen based on data from humans as well as the robotSpring Flamingo and preliminary dynamic computer simulations.
4.2 The Design
4.2.1 Overall Structure
A photo of the robot and a joint schematic view are shown in Figure 4-1. The leg of the robot hassix active degrees of freedom plus an optional passive degree of freedom in the foot. The verticalaxis, Z, is the yaw axis. The X axis is the roll axis and the Y axis is the pitch axis. The hip hasthree degrees of freedom. These three degrees of freedom are made up of a universal joint (yaw androll) followed by a pin joint (pitch). The pitch pin joint is offset slightly(about 2cm) from the yawand roll axes.
The frontal plane dimensions of the robot are shown in Figure 4-2. The dimensions are very closeto the dimensions for a 50th percentile US male as given by Whitney(2). The ranges of motion areadapted from robot simulations and data found in Rose, et al.(18), Winter(20), and Kapandji(7).
deg rad/s Nm Drive Typehip(pitch) 80,-30 7.3333 50 Pulleyhip(roll) 30,-20 6.8 59 Push-rodhip(yaw) 30,-15 5.5 67 Push-rod
knee(pitch) 80,0 8.8 42 Push-rodankle(pitch) 45,-20 8.8 88* Push-rodankle(roll) 20,-20 7.3 100* Push-rod
Table 4.1: Robot Joint Specifications. Torque and rad/s numbers are given for maximum momentarm. Power, torque, and velocity are symmetric due to the actuator. *The ankle roll and pitch arenot independent. Their maximum values can not be applied simultaneously.
19
CHAPTER 4. ROBOT MECHANICAL DESIGN 20
X - rollY - pitch
Z - yaw
Sagittal
Frontal
Transverse
Figure 4-1: A photo of the completed robot and a joint schematic view of the robot. The schematicshows active degrees of freedom only. The optional passive toe joint is not shown.
CHAPTER 4. ROBOT MECHANICAL DESIGN 21
0
3.00
20.00
37.00
40.00
60.00
(Hip Spacing)7.25
16.00
Shin and Thigh Width
4.50
4.00
37.75
Ankle Universal(Pitch, Roll)
Knee Pitch Axis
Hip Pitch Axis
Hip Universal(Yaw, Roll)
Figure 4-2: The dimensions of the biped in the frontal plane. The body and links are approximatelyaxially symmetric about their longitudinal axis. Dimensions are in inches.
CHAPTER 4. ROBOT MECHANICAL DESIGN 22
Body Part/Area Human Biped Robot PDWShin & Foot (x2) 6% 13.5% 10%
Thigh (x2) 10% 11% 15%Ab/Pelvic 27% 51 % 50%Arm (x2) 5% NA NA
Thorax to Head 31% NA NA
Table 4.2: Approximate distribution of mass in humans, the biped M2, and McGeer’s kneed passivedynamic walker(PDW). Human data adapted from Dempster and Gaughran (3). Robot weightdistribution is driven primarily by actuator locations. Each actuator is approximately 1.2kg(2.5lbs).
The mass distribution of the robot is dominated by the location of the actuators within thelinks. As a result, the robot’s mass distribution is centered lower than an average human’s. Table4.2 shows the percentage mass distributions for an average male, M2, and a planar passive dynamicwalker(10). The robot mass distribution is closer to that of a planar passive dynamic walker than tothat of a human. Due to successful computer simulations(16), it was not necessary to add additionalweight to the torso in order to put the proportions more in line with a human.
4.2.2 Actuators
The actuators used in the robot are 90W, 1.2KG Series Elastic Actuators(17). They are capable ofa maximum force of roughly 1320N (300lbs) and a maximum speed of roughly 0.28 m/s (11 in/s).The actuators have a force control bandwidth of 30Hz. Linear actuators where chosen over rotaryactuators due to the available space in the robot. Linear actuators allowed for placement along thelongitudinal axis of the leg links. The actuators are symmetric in their power, speed, and forcecapabilities.
4.2.3 Ankle
The ankle of the biped is a universal joint. The pitch axis is followed by the roll axis. This is a slightdeviation from the structure of the human ankle. The human ankle is often likened to a universaljoint where the second axis is at 45 degrees to the first rather than at 90 degrees(7). For engineeringsimplicity we use a universal joint with orthogonal axes. The instantaneous power requirement forankle pitch is the greatest. The actuators were placed in a configuration so they can act together inthe pitch direction.
Photos of the ankle are shown in Figure 4-7 and a schematic of the ankle and its actuators areshown in Figure 4-3. The ankle has two series elastic actuators placed along the longitudinal axisof the shin. The actuators are mounted by a universal joint near the top of the shin and attachedto the foot by a ball and socket joint (rod-end). This is a linkage variation of a standard geareddifferential. When the actuators push in unison, a moment is generated about the pitch axis. Whenthey push in opposite directions, a roll moment is generated.
4.2.4 Foot
Two different feet were designed for the robot. One foot is a simple rectangular design with foursingle axis load cells residing at each of the corners. This foot closely resembles the foot that wasused in computer simulations in the lab. Another more involved foot was designed in order to explorethe roll of the toes in walking.
CHAPTER 4. ROBOT MECHANICAL DESIGN 23
X axis
Y axis
Ball Joint
Actuators
E
A
O
Figure 4-3: A schematic of the ankle joint actuation scheme. The axes shown are fixed to the centerof the universal joint. The points A,E, and O are referenced in Appendix A.
2.88
0
2.13
9.59
.50
6.69
9.09
.19
4.00AnkleJoint
ToeJointAxis
Heel
Figure 4-4: The dimensions of the biped foot with passive toe joint. The X’s represent the locationof the load cells.
The more intricate second foot of the biped robot contains a passive joint which is modeled afterthe toe of a human. The joint is believed to smooth the center of mass trajectory of the body duringa walking cycle(18). The toe joint is simply a pin joint with two limit stops and a soft return spring.
Ground contact and sensing on the foot consists of four single axis load cells. One cell is placedat the heel and three cells are placed in a triangle at the toe/ball of the foot. The three cells in thetoe are all constrained to the same plane. The three toe contact points can rotate about the footY-axis with respect to the heel contact point. The layout of the sensors can be seen in Figure 4-4.
Standard through-hole button load cells are used. There is a rubber bumper which contacts theground roughly 0.875” in diameter attached to each load cell. On the simple four point foot thereis a rectangular pastic piece with a 0.25” piece of neoprene for grip and shock absorbtion. This canbe seen in Figure 4-7. Single axis load cells were chosen over a six axis sensor because of their sizeand weight, and chosen over strain gauges because of their ease of use and quick replaceablility.
On the more complicated biped foot, force control and a passive toe joint are the reasons thefour contact points of the foot are not over-constrained. Since ankle roll is force controlled ratherthan position controlled, it can adjust itself so the three points of the toe lie flat. The passive toe
CHAPTER 4. ROBOT MECHANICAL DESIGN 24
Body frame
Yaw
Roll
Thigh
Pitch
Yaw
ROLL
O
Mr
ArMy
Figure 4-5: A schematic of the biped hip. The pitch actuator is not shown. It lies along the axis ofthe thigh.
Figure 4-6: A photo of the biped thigh illustrating the hip pitch cable drive system.
joint(a pitch joint) then allows the fourth point on the heel to lie flat as well.
4.2.5 Hip
The biped hip has three degrees of freedom. The joint consists of a universal joint followed by aslightly offset pin joint. A schematic of the hip is shown in Figure 4-5. The yaw axis is first and thethe yaw actuator is mounted to the body frame with a pin joint. The roll actuator is next. Since itsattachment point passes through the roll angle, it is mounted to the body by a universal joint. Itsendpoint is attached by a ball and socket joint. The pitch actuator, which is not shown, lies alongthe longitudinal axis of the thigh. It is the only actuator which is attached using a cable and pulleyrather than a rod-end. The range of motion of the hip pitch joint is the largest. The moment armchanges associated with a drive arm would be too great at the extents of the pitch motion. At sixtydegrees from the perpendicular drive position, the moment arm would be half its original length.
CHAPTER 4. ROBOT MECHANICAL DESIGN 25
45 deg20 deg
20 deg
Figure 4-7: The ranges of motion on the robot ankle joint. The ankle roll is symmetric(+- 20 deg).
4.2.6 Knee
The robot knee is a simple pin joint. The joint is actuated by a SEA located in the thigh. Twophotos demonstrating the range of motion of the knee can be seen in Figure 4-8.
CHAPTER 4. ROBOT MECHANICAL DESIGN 26
80 deg
Figure 4-8: The range of motion of the robot knee. The knee has 80 degrees of motion.
Chapter 5
Electronics and Control Systems
5.1 Electronics Overview
The basic electronic subsystems of the robot are shown in Figure 5-1. The robot is powered by a48 Volt power supply. There is also an Ethernet connection not shown in the figure which can beused to load control code and retrieve data from the on-board computer. The circuit boards aresummarized in Table 5.1. The sensors are summarized in Table 5.2.
All signals are sent and received differentially between the computer and the analog boards.
5.1.1 48 Volt bus
All the electronics on the robot receive power from a 48 Volt bus. The current setup has the robotconnected to several offboard 48V power supplies. All the electronics were designed with this powerbus in mind. The voltage choice was dictated by the brushless motor amplifiers. It was also chosenbecause it is a standard voltage and the robot will eventually be augmented with batteries for futureautonomous demos.
5.1.2 Robot Sensors
The robot is equipped with a total of 33 sensors. There are 12 rotary potentiometers that are usedto measure joint position and this is differentiated to give the joint velocity. The joint positionmeasurements are relative displacements between joints. There are 12 linear potentiometers whichmeasure the spring compression in the Series Elastic Actuators. These signals are local to the analogforce control loop only. They are not used by the DSP in the control. There are 8 single axis loadcells on the feet of the robot. These sensors are connected to the strain gauge conditioning boardwhich amplifies and buffers the signal on its way to the DSP. There is a single Intersense Inertia
Table 5.1: Circuit Boards on the Robot
QTY Voltage Input VendorMotor Amplifiers 12 +48 Copley Controls
DSP 1 +5 DideasANA074 1 5,12 DideasANA063 1 5,12 Dideas
Analog Force Control Board 6 +48 Leg LabPower Distribution Board 2 +48 Leg LabSignal Conditioning Board 4 12 Leg LabIS-300 Signal Conditioner 1 +12 Intersense
27
CHAPTER 5. ELECTRONICS AND CONTROL SYSTEMS 28
Table 5.2: Sensors on the Robot
Sensor QTY Voltage Input VendorRotary Potentiometers 12 +-5 BournsLinear Potentiometers 12 +-5 NovotechnikSingle-Axis Load Cells 8 +-9 Transducer Techniques
3-axis inclinometer 1 NA Intersense
Power+-5+-12
Instrumentation
Load Cells
ActuatorMotor AmpBrushless
A/D and D/A
Computer
Regulated 5VAnalog PID
Buffers
JointPots
Vestibular Sensors
48Volt bus
Figure 5-1: An overview of the biped electronics. Thick lines indicate power transfer and arrowheadsindicate information flow.
CHAPTER 5. ELECTRONICS AND CONTROL SYSTEMS 29
Cube which measure the roll, pitch, and yaw(magnetic) of the robot body. The Inertia Cube isconnected to a IS-300 motion tracker unit which communicates with the DSP through a RS-232serial line. Table 5.2 provides a summary of the robot sensors. Figure 5-2 contains an annotatedphoto of M2 with the sensor locations.
Figure 5-2: A photo of M2 with annotations of electrical systems.
5.2 Custom Circuit Boards
5.2.1 The Signal Conditioning Boards
The signal conditioning boards are four layer printed circuit boards. They have a ground plane, apower plane, and two routing layers. The signal conditioning boards handle two 350 ohm wheatstonebridge sensors each. They have adjustable gain and offsets.
5.2.2 The Analog Force Control Boards
The analog force control board provides two channels of PID force control for the Series ElasticActuators and two channels of joint potentiometer buffering and differentiation. The analog forcecontrol boards have a 48 Volt power input and communicate with the DSP through differentialamplifiers and receivers.
The analog force control boards are four layer printed circuit boards. They have a ground plane,a power plane, and two routing layers.
CHAPTER 5. ELECTRONICS AND CONTROL SYSTEMS 30
Figure 5-3: A photo of the biped strain gauge signal conditioning board. This board handles twostrain gauges. There are a total of four of these boards on the robot.
Figure 5-4: A photo of the analog force control and joint pot buffer board. There are six identicalboards on the robot.
CHAPTER 5. ELECTRONICS AND CONTROL SYSTEMS 31
5.2.3 The Analog Breakout Board
The analog breakout board connects to the ANA070 and ANA064 analog input and output boards.It has no power and simply allows for the use of smaller connectors to easily distribute the 64channels of analog input and output from the ANA070 and ANA064 boards. There are two 64 pinconnectors which connect to the ANA064 and ANA070 boards respectively. Then there are multiple3,5, and 10 pin connectors which connect to the signal conditioning boards and analog force controlboards. The analog breakout board is a two layer printed circuit board.
5.2.4 The Power Distribution Boards
The power distribution board has a 48 Volt input and outputs a variety of lower voltages. It appearsin two places on the robot. There is one printed circuit for the two power distribution boards butthere are two different DC-to-DC converters which are used when assembling the board.
• Board Configuration 1 This board supplies power to the DSP (+5V) and two analog boards(+-12V, +-5V). The analog and digital grounds are isolated.
• Board Configuration 2 This board provides power to the Intersense IS-300(+12V) and thefour signal conditioning boards (+-12V).
The power distribution boards are four layer printed circuit boards. They have a ground plane,a power plane, and two routing layers.
5.2.5 The DSP and Analog I/O
M2 has a custom DSP system that was designed by Chris Barnhart at Digital Designs and Systems.The DSP is a Texas Instruments C-31. There are three circuits boards that make up the system.
1. DSP. Mulitple Serial Lines and Digital I/O.
2. 16 Analog Ins, 16 Analog Outs, and Ethernet.
3. 32 Analog Inputs.
5.3 Control Software
Control code for the robot is written in C. It is compiled offline on a Unix machine and downloadedto the robot via Ethernet. The main robot control code runs at 500Hz on the DSP and sends desiredforces to the analog boards. Currently, the control system treats the actuator and analog boards asa ’black box’ force source. This means it is assumed that the desired force sent out to the actuatoris the actual force applied to the joint. This is true within the realm of frequencies needed forwalking(less than 15Hz). The analog boards have PD control loops for the Series Elastic Actuators.The analog boards also buffer and differentiate the joint potentiometers and send the informationto the DSP.
This is similar to the electronics used for Spring Flamingo. The main differences are the additionof Ethernet for faster communication with the outside world and differential send and receive forsignals within the robot.
Bibliography
[1] Jesper Adolfsson, Harry Dankowicz, and Arne Nordmark. 3-d stable gait in passive bipedalmechanisms. Proceedings of 357 Euromech, 1998.
[2] Henry Dreyfus Associates. The Measure of Man and Woman. Whitney Library of Design, NewYork, 1993.
[3] W.T. Dempster and G. Gaughran. Properties of body segments based on size and weight.American Journal of Anatomy, 1965.
[4] J. Fowble and A. Kuo. Stability and control of passive locomotion in 3d. Proceedings of theConference on Biomechanics and Neural Control of Movement, pages 28–29, 1996.
[5] Mariano Garcia, Anindya Chatterjee, and Andy Ruina. Speed, efficiency, and stability ofsmall-slope 2d passive dynamic bipedal walking. IEEE International Conference on Roboticsand Automation, pages 2351–2356, 1998.
[6] K. Hirai, M. Hirose, Y. Haikawa, and T. Takenaka. The development of honda humanoid robot.IEEE International Conference on Robotics and Automation, 1998.
[7] Ibrahim Adalbert Kapandji. The Physiology of the Joints. Churchill Livingstone, 1982.
[8] Benjamin Krupp. Design and control of a planar robot to study quadrupedal locomotion.Master’s thesis, Massachusetts Institute of Technology, August 2000.
[9] Tad McGeer. Passive dynamic walking. International Journal of Robotics Research, 9(2):62–82,1990.
[10] Tad McGeer. Passive dynamic biped catalogue. Proceedings of the 2nd InternationalSymposium of Experimental Robotics, 1991.
[11] Thomas A. McMahon. Mechanics of locomotion. The International Journal of RoboticsReasearch, 3(2):4–28, 1984.
[12] Gill A. Pratt and Matthew M. Williamson. Series elastic actuators. IEEE InternationalConference on Intelligent Robots and Systems, 1:399–406, 1995.
[13] J. Pratt, P. Dilworth, and G. Pratt. Virtual model control of a bipedal walking robot. IEEEInternational Conference on Robotics and Automation, pages 193–198, 1997.
[14] Jerry E. Pratt. Virtual model control of a biped walking robot. Master’s thesis, MassachusettsInstitute of Technology, August 1995.
[15] Jerry E. Pratt and Gill A. Pratt. Exploiting natural dynamics in the control of a planar bipedalwalking robot. Proceedings of the Thirty-Sixth Annual Allerton Conference on Communication,Control, and Computing, pages 739–748, 1998.
[16] Jerry E. Pratt and Gill A. Pratt. Exploiting natural dynamics in the control of a 3d bipedalwalking simulation. Proceedings of the International Conference on Climbing and WalkingRobots (CLAWAR99), 1999.
32
BIBLIOGRAPHY 33
[17] David W. Robinson, Jerry E. Pratt, Daniel J. Paluska, and Gill A. Pratt. Series elastic actua-tor development for a biomimetic robot. IEEE/ASME International Conference on AdvancedIntelligent Mechatronics, 1999.
[18] Jessica Rose and James G. Gamble. Human Walking. Williams and Wilkins, 1994.
[19] Mark Roshiem. Robot Wrist Actuators. John Wiley and Sons, Inc., 1989.
[20] D. A. Winter. Biomechanics and Motor Control of Human Movement. John Wiley and Sons,Inc., New York, 1990.
[21] Jinichi Yamaguchi, Eiji Soga, Sadatoshi Inoue, and Atsuo Takanishi. Development of a bipedalhumanoid robot - control method of whole body cooperative dynamic biped walking. IEEEInternational Conference on Robotics and Automation, pages 368–374, 1999.
Appendix A
Joint Math
Since all of M2’s rotary joints are actuated by linear actuators, transformations are needed to getfrom joint torques to actuator forces. These are dependent upon the joint layout and actuationscheme. If the force is known, the torque is somewhat easily computed using the formula, τ = �r× �F .
The cross product τ = �r × �F operation can be rewritten as a matrix multiplication τ = A �Fwhere A is a singular matrix.
τ = �r × �F =
∣∣∣∣∣∣i j kr1 r2 r3
F1 F2 F3
∣∣∣∣∣∣=
0 −r3 r2
r3 0 −r1
−r2 r1 0
�F = A�F (A.1)
However, if the cross-product is written as τ = |�R||�F | sin(θ) then we can easily invert. In thecase of most of the robot joints, this requires use of the arctan function or the law of cosines tofind the angle θ. In general, the direction of �F is fixed by the geometry of the joint and |�F | is thequantity of real interest.
A.1 Knee Joint
The knee joint is a single degree of freedom rotary joint allowing relative motion between the shinand the thigh. When all the joints of the robot are zero(it is standing with locked knees), the kneeaxis is parallel to the global Y axis. Rotation about the knee joint is referred to as pitch.
The knee actuator is connected to the knee joint via a push rod of length | �OA| = rk at the shinand a fixed pin joint at the thigh.
The knee joint has three points of interest which we will use for the derivation of the transfor-mation. The knee pivot O, the actuator pushrod attachment A, and the actuator mounting pivotM . The robot knee joint and points can be seen in Figure A-1 and a simple line drawing is seen inFigure A-2.
τ = �r × �F = rk sin(� OAM)Fknee (A.2)
The actuator force required given τk is
Fknee =τk
rk sin(� OAM)(A.3)
where Fknee is a scalar and � OAM can be defined as follows
� OAM = θfixed + θk − � AM = θfixed + θk − arctanr sin(θfixed + θk) − L2
L1 + r sin(θfixed + θk)(A.4)
34
APPENDIX A. JOINT MATH 35
Figure A-1: The robot knee joint with superimposed lines and points. Points O, A and M all referto pin joints. O is the knee joint. M is where the actuator is mounted to the thigh and A is wherethe actuator is attached to the shin.
θfixed
θknee
O
AM
L2
L1
MA
Figure A-2: A line drawing of the knee for the calculation of knee actuator desired force. Theactuator is the line segment �MA. This drawing also pertains to the geometry of the hip and anklejoints but the hip and ankle actuators have motions out of the plane whereas the knee actuator isalways in the plane perpendicular to the knee axis.
APPENDIX A. JOINT MATH 36
X
Y O
A1A2
rx
ry
r
Figure A-3: A simple view of the actuation of the robot ankle. This is a tranverse plane slice ofthe robot ankle. Point O is the intersection of the X and Y axes as well as the center of the ankleuniversal joint. Points A1 and A2 represent the attachment points of the two actuators respectively.Forces in the actuators create torques about both the X and Y axes. Joint rotations change therelative lengths of the X and Y projections(rx and ry) of the moment arms, therefore affecting thetorque to force transformation.
The angle between the shin and the thigh is θk and the constant, θfixed, is the angle between theshin and the segment �OA. It is also possible to derive the equations avoiding the inverse tangent byusing the law of cosines.
It would be possible to simplify the equations a bit by assuming the actuator is always parallelto the thigh, i.e. � �AM = 0. This simplifies the equations and shouldn’t reduce accuracy muchconsidering that L1 � L2. Equation A.3 then becomes
Fknee =τk
rk sin(θfixed + θk)(A.5)
where Fknee is once again a scalar quantity.
A.2 Ankle Joint
The ankle joint is a universal joint(2 d.o.f.) allowing relative motion between the foot and the shin.When all of the joints of the robot are zero, the axes of the ankle joint line up with the global Xand Y axes. The Y axis is proximal to the shin. Rotation about the Y axis is referred to as pitchand rotation about the X axis is referred to as roll. The math for the ankle is an extension of themath for the knee.
We can consider the ankle as two decoupled cases of the knee joint. Then one can simply addthe forces from τaroll and τapitch. There is a slightly more complicated analogy to the knee in twoplanes of the ankle. The ankle roll considers what is happening in the XZ plane and the ankle pitchconsiders what is happening in the YZ plane. Both planes are coming out of the page in FigureA-3. Each plane looks similar to the knee diagram shown in Figure A-2. The main difference is that|OA| = r is a function of the other ankle angle.
APPENDIX A. JOINT MATH 37
Body frame
Yaw
Roll
Thigh
Pitch
Yaw
ROLL
O
Mr
ArMy
Figure A-4: The robot hip joint schematic. Point O refers to the hip universal joint. The Z(hipyaw) and Y(hip roll) axes intersect at point O. Point Mr refers to the universal joint on the bodywhere the roll actuator is mounted. Point Ar refers to the ball joint on the thigh where the rollactuator is attached. Point My refers to the pin joint where the yaw actuator is attached to body.Point Ay is the pin joint attachment of the yaw actuator to the yaw universal block. It is hidden inthis schematic.
A.3 Hip Joint
A.3.1 Hip Pitch
The math for the hip pitch is the simplest of all the joints since the hip pitch actuator is attachedby a pulley. The desired torque at the joint(τhp), leads to the force(fhp) command to the hip pitchlinear actuator.
fhp =τhp
rhippitch(A.6)
The transformation from actuator velocity to joint velocity is quite simple as well.
ωhp =velhp
2πrhp(A.7)
A.3.2 Hip Roll and Yaw
The yaw actuator transformation is identical to that of the knee actuator except for the addition ofa small term which can be ignored with little affect.
The actuator force required given τhyaw is
Fhyaw =τhyaw
sin(� OyAyMy)+ εhroll (A.8)
where ε is a small factor due to the torque applied to hip roll which is a function of the hip rolland yaw angles. Oy, Ay, and My are defined as they were in the case of the knee.
APPENDIX A. JOINT MATH 38
The actuator force required given τhroll can be simplified to
Fhroll =τhroll
sin(� OhrAhrMhr)(1 + εhyaw) (A.9)
The exact equation for the hip roll incorporates the yaw angle. Since the hip roll actuatorattachment point Ahr is attached to the leg below the yaw degree of freedom, the points Ohr, Ahr,and Mhr do not always lie within the plane perpendicular to the roll axis as in the case of the kneejoint where they are fixed in the plane perpendicular to the knee axis.
When the roll actuation plane(plane formed by the three points Ohr, Ahr, Mhr) is skewed withrespect to the roll axis, a small torque will be contributed to the yaw axis. This is where the termεhroll comes from in Equation A.8 and this is why the hip roll equation has the εhyaw term in it.
Appendix B
Electrical Schematics
This Appendix contains the electrical schematics for all the custom circuit boards used on the robot.The basic functionality of the boards and their role in the overall robot is decribed in more detailin Chapter 5. There are six pages for the analog force board, 2 pages for the signal conditioningboard, 1 page for the analog breakout board, and 1 page for the power distribution board.
39
APPENDIX B. ELECTRICAL SCHEMATICS 40
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
2 Actuator Command from DSP
2 Current Commands on their
way to Copleys
2 Joint Pot Signals out to
DSP
2 Joint Pot Derivatives
out to DSP
PO
WE
R A
ND
CO
NN
EC
TOR
S{R
evC
ode}
MIT
LE
G L
AB
, BIP
ED
FO
RC
E B
OA
RD
A
16
Tue
sday
, Dec
embe
r 21
, 199
9
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
+5
-5
+5
Join
t_P
ot_B
-5
Join
t_P
ot_A+5
+5
-5
+12
Com
man
d_In
_A+
Com
man
d_In
_A-
Com
man
d_In
_B+
Com
man
d_In
_B-
Cop
ley_
B-
Cop
ley_
B+
Loca
l_G
ND
Cop
ley_
A-
Cop
ley_
A+
-12
+12
-12
Loca
l_G
ND
Join
t_P
ot_B
-_ou
t
Loca
l_G
ND
Join
t_V
el_B
-_ou
t
Join
t_P
ot_A
+_o
ut
Join
t_P
ot_A
-_ou
t
Join
t_V
el_A
+_o
ut
Join
t_V
el_A
-_ou
t
Join
t_P
ot_B
+_o
ut
Join
t_V
el_B
+_o
ut
For
ce_P
ot_B
For
ce_P
ot_A
+5
-5
Loca
l_G
ND
-5
12-
12+
CC
B-
CC
B+
CB
-
CB
+
CA
-
CA
+
CC
A-
CC
A+
5-
5+
FP
B
FP
AJP
A
JPB
JVB
-
JVB
+
JVA
-
JVA
+
JPB
-
JPB
+
JPA
-
JPA
+
J1 48IN
_CO
NN1 2
+48
GN
DDC/DC
CONVERTER
+-12V
PB
1 DF
C10
U48
D12
1 2
3 4 5
+IN
PU
T
-IN
PU
T
+12O
UT
CM
N
-12O
UT
J2
4pin
1 2 3 4
+Vo
A
-Vo
A
+Vo
B
-Vo
B
J4
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
TP
4m
inus
5
Test Point
TP
3
plus
5
Tes
t Poi
nt
TP
2m
inus
12
Test Point
GN
D1
GN
D
Test Point
TP
1
plus
12
Tes
t Poi
nt
PC
210
uf
1 2
PC
3
10uf
12
PC
122
0uf
1 2
J3
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
BC
11u
f
1 2
BC
21u
f
1 2
GN
D2
GN
D2
Test Point
RE
G2
I433
14
13
2
GN
D-5
OU
T
IN
RE
G1
I434
06
13
2
IN+5
OU
T
GND
GN
D3
GN
D0
Test Point
GN
D4
GN
D20
Test Point
JFP
2 4pin
_pot
1 2 3 4
+5 Sig
nal
-5 GN
D
JJP
1
4pin
_pot
1 2 3 4
+5 Sig
nal
-5 GN
D
JJP
2
4pin
_pot
1 2 3 4
+5 Sig
nal
-5 GN
D
JFP
1
4pin
_pot
1 2 3 4
+5 Sig
nal
-5 GN
D
Figure B-1: Page 1 of 6 of the biped force board.
APPENDIX B. ELECTRICAL SCHEMATICS 41
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
FOR
CE
ER
RO
R, A
and
B{R
evC
ode}
MIT
LE
G L
AB
, BIP
ED
FO
RC
E B
OA
RD
A
26
Wed
nesd
ay, D
ecem
ber
15, 1
999
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
Loca
l_G
ND
Loca
l_G
ND
Loca
l_G
ND
Loca
l_G
ND
FP
B
FP
A
FE
B
FE
A
CA
-
CA
+
CB
-
CB
+
FA
_DE
S
FB
_DE
S
5-5+
5+
5-
12+
12+
12+
12-
12-
12-
PO
T2
20k
1 3
2
PO
T1
20k
1 3
2
R4
10k
R7
10k
R16 5k
R14 5k
R3 5k
R5 5k
R11
10k
R15
10k
R2 5k
R13 5k
R10 5k
R8 5k R
12
5k
R1
5k
R9
5k
R6
5k
G=1/2
RC
V1
INA
2137
2 3
4
56
8910
11
12 13 14
-In
A
+IN
A V-
+In
B
-In
B
Ref
B
Out
B
Sen
se B
V+
Sen
se A
Out
A
Ref
A
TP
8 Test Point TP
9Test Point
TP
6 Test Point
TP
10Test Point
TP
5
Tes
t Poi
nt
TP
12
Tes
t Poi
nt
TP
7
Tes
t Poi
nt
TP
11
Tes
t Poi
nt
+-
U1D
LM32
41213
14
411
+-
U2A
LM32
432
1
411
+ -
U1C
LM32
4
10 98
4 11
+ -
U2B
LM32
4
5 67
4 11
+-
U2C
LM32
4109
8
411
+-
U1B
LM32
456
7
411
+-
U1A
LM32
432
1
411
+-
U2D
LM32
41213
14
411
BC
4
1uf
1 2
BC
8
0.1u
f
1 2BC
5
0.1u
f
1 2
BC
7
1uf
12
BC
30.
1uf
1 2
BC
60.
1uf
1 2
FC
1 0.1u
f
FC
2 0.1u
f
Figure B-2: Page 2 of 6 the biped force board.
APPENDIX B. ELECTRICAL SCHEMATICS 42
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
PID
A{R
evC
ode}
MIT
LE
G L
AB
, BIP
ED
FO
RC
E B
OA
RD
A
36
Wed
nesd
ay, D
ecem
ber
15, 1
999
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
CC
A+
CC
A-
FE
A
12-
12+
12+
12-
FA
_DE
S
R17 5k
R20 40
.2 k
R18
5k
C3 op
en
C2
1uf
C1
.01u
f
R22
402
ohm
R25
open
R19
357
ohm
R24 op
en
R26
open
R23
4.42
K
R21 10
k
+-
U3A
LM32
4_0
321
411
+-
U3B
LM32
4_0
567
411
+-
U3C
LM32
4_0
1098
411
DR
V1
DR
V13
4
1234
56
78
-Vo
-Sen
seG
nd
Vin
V-V+
+S
ense
+Vo
+-
U3D
LM32
4_0
121314
411
TP
13
PID
_A
Test Point
BC
9
0.1u
f
12
BC
12
0.1u
f
12
BC
10
0.1u
f12
BC
11
0.1u
f
12
RF
F1
Ope
n
Figure B-3: Page 3 of 6 of the biped force board.
APPENDIX B. ELECTRICAL SCHEMATICS 43
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
PID
B{R
evC
ode}
MIT
LE
G L
AB
, BIP
ED
FO
RC
E B
OA
RD
A
46
Wed
nesd
ay, D
ecem
ber
15, 1
999
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
ofCC
B+
CC
B-
FE
B
12-
12+
12+
12-
FB
_DE
S
R27 5k
R30 40
.2k
R28
5k
C6
open
C5
1uf
C4
.01u
f
R32
402
ohm
R35
open
R29
357
ohm
R34
open
R36
open
R33
4.42
k
R31 10
k
+-
U4A
LM32
4_0
321
411
+-
U4B
LM32
4_0
567
411
+-
U4C
LM32
4_0
1098
411
DR
V2
DR
V13
4
1234
56
78
-Vo
-Sen
seG
nd
Vin
V-V+
+S
ense
+Vo
+-
U4D
LM32
4_0
121314
411
TP
14
PID
_B
Test Point
BC
14 0.1u
f
12
BC
15 0.1u
f1
2
BC
160.
1uf
12
BC
13 0.1u
f1
2
RF
F2
open
Figure B-4: Page 4 of 6 the biped force board.
APPENDIX B. ELECTRICAL SCHEMATICS 44
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
JOIN
T P
OS
ITIO
N A
ND
VE
LOC
ITY
{Rev
Cod
e}
MIT
LE
G L
AB
, BIP
ED
FO
RC
E B
OA
RD
A
56
Wed
nesd
ay, D
ecem
ber
15, 1
999
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
Loca
l_G
ND
JPA JP
B
JPA
+
JPA
-
JVA
+
JVA
- JPB
+
JPB
-
JVB
+
JVB
-
12+
12-
12+
12-
12-
12+
12+
12-
12-
12+
12+
12-
JPB
_BU
F
JPA
_BU
F
JPB
_BU
F
JPA
_BU
FR
38 10k
R43 10
k
R42
5k
C10
0.03
3uf
C7 0.
033u
fR
375k
R39
5k
C8
0.33
uf
R44
5k
C11
0.33
uf
C12
1uF
C9
1uF
R40 1.
65 k
R45 1.
65 k
TP
17
Test Point
TP
18
Test Point
TP
16
Test Point
TP
15
Tes
t Poi
nt+ -
U5A
LM32
4
3 21
4 11
+ -
U6A
LM32
4
3 21
4 11
+-
U5B
LM32
456
7
411
+-
U6B
LM32
456
7
411
+-
U5C
LM32
4109
8
411 +-
U6C
LM32
4109
8
411
+-
U5D
LM32
4
121314
411+-U
6D LM32
41213
14
411
DR
V3
DR
V13
4
1234
56
78
-Vo
-Sen
seG
nd
Vin
V-V+
+S
ense
+Vo
DR
V5
DR
V13
4
1234
56
78
-Vo
-Sen
seG
nd
Vin
V-V+
+S
ense
+Vo
DR
V4
DR
V13
4
1234
56
78
-Vo
-Sen
seG
nd
Vin
V-V+
+S
ense
+Vo
DR
V6
DR
V13
4
1234
56
78
-Vo
-Sen
seG
nd
Vin
V-V+
+S
ense
+Vo
D4 LE
D
D3 LE
D
D1 LE
D
D2 LE
D
R46 1k
R41 1k
BC
17
0.1u
f
12
BC
19 0.1u
f
12
BC
23
0.1u
f
12
BC
24 0.1u
f1
2
BC
25
0.1u
f
12
BC
28
0.1u
f1
2
BC
22
0.1u
f12
BC
21
0.1u
f
12
BC
26
0.1u
f1
2B
C27
0.1u
f1
2
BC
20
0.1u
f
12
BC
18
0.1u
f
12
RL3 50
k
RL2 50
k
RL1
5k
RL4 5k
Figure B-5: Page 5 of 6 of the biped force board.
APPENDIX B. ELECTRICAL SCHEMATICS 45
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
LED
MO
NIT
OR
S{R
evC
ode}
MIT
LE
G L
AB
OR
AT
OR
Y, B
IPE
D F
OR
CE
BO
AR
D
A
66
Wed
nesd
ay, D
ecem
ber
15, 1
999
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
12+
12-
FB
_DE
S
FA
_DE
S FE
A
FE
B
D8 LE
D
D7 LE
DR
48 500o
hm
+-
U7A
LM32
4
321
411
D6 LE
D
D5 LE
D
R47 50
0ohm
RL5 50
K
RL6
5k
D9 LE
D
R49 1
k
R50 1
kD
12 LED
D11 LE
D
D10 LE
D
BC
29
0.1u
f
12
BC
30
0.1u
f
12
JMP
1
JUM
PE
R+-
U7D
LM32
4
121314
411RL1
1
open
RL1
2
5k
JMP
4
JUM
PE
R
+-
U7C
LM32
4
1098
411
RL9 op
en
RL1
0
5k
JMP
3
JUM
PE
R+-
U7B
LM32
4
567
411RL7 50
K
RL8 5k
JMP
2
JUM
PE
R
Figure B-6: Page 6 of 6 the biped force board.
APPENDIX B. ELECTRICAL SCHEMATICS 46
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
StrainGainAdjust
For Half-Bridge Completion:
1 - Connect pin 2 and pin 25
2 - Leave connector -Input open
MIT Leg Laboratory
M2/
M4
Pro
topt
ype
Str
ain
Gag
e B
oard
A
12
Frid
ay, A
ugus
t 14,
199
8
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
ofStrain 2
Strain 1
-Inp
ut 1
Str
ain
1
+Inp
ut 1
Sen
se-H
igh
1V
exc_
out 1
Sen
se-L
ow 1
-Inp
ut 2
Sen
se-L
ow 2
Sen
se-H
igh
2V
exc_
out 2
+Inp
ut 2
Str
ain
2
+12V
+12V
-12V
-12V
-12V
-12V
+12V
+9V
reg
-9V
reg
-9V
reg
-9V
reg
-9V
reg
-12V
+9V
reg
+12V
+9V
reg
+9V
reg
+12V
-12V
+12V
R1
Str
ain
Gai
n A
djus
t 1
R2
15k
TP
1 Test Point
TP
2 Test Point
TP
3 Test Point
TP
6 Test Point
TP
4 Test Point
TP
7 Test Point
TP
8 Test Point
R4
620
C3
1uF
J4 ST
RA
IN_C
ON
N
1 2S
trai
nG
ND
TP
5 Test Point
C2
1uF
U4
I021
64
13
2
GN
D-9
OU
T
IN
J1 BR
IDG
E_C
ON
N
1 2 4 5 63
Vex
c_ou
tS
ense
-Hig
h
-Inp
utS
ense
-Low
GN
D
+Inp
ut
C1
4.7u
F
J2 ST
RA
IN_C
ON
N
1 2S
trai
nG
ND
J5
FLA
M_P
WR
_CO
NN
2 31
GN
D-1
2V
+12V
J3 BR
IDG
E_C
ON
N
1 2 4 5 63
Vex
c_ou
tS
ense
-Hig
h
-Inp
utS
ense
-Low
GN
D
+Inp
ut
C4
1000
pFC
510
00pF
D4 LE
D
2 1
D3 LE
D
2 1
D1 LE
D
2 1
TP
9 Test Point
R3
50k
1 3
2
D2 LE
D
2 1
R5
470
R6
470
U3
I021
62
13
2
IN+9
OU
T
GND
R7
620
C6 0.
1uF
+-
U1A
LF34
7
321
411
Strain Gage
Signal Conditioner
U2
1B31
1 2 3 4 8 9 10 11 12 13 141516171819202125262728
+ In
put
- In
put
Gai
n 1
Gai
n 2
Vou
t Unf
ilter
edIn
put O
ffset
Adj
1In
put O
ffset
Adj
2O
utpu
t Offs
et A
djB
andw
idth
Adj
. 1B
andw
idth
Adj
. 2V
out F
ilter
ed-
Vs
Com
m+
Vs
+ V
s R
egR
ef. O
utR
ef. I
nE
xc. A
dj
Hal
f-B
ridge
Com
plet
ion
Sen
se L
owS
ense
Hig
hV
exc
Out
C7
0.1u
F
Figure B-7: Page 1 of 2 of the strain gauge conditioning board.
APPENDIX B. ELECTRICAL SCHEMATICS 47
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
StrainGainAdjust
For Half-Bridge Completion:
1 - Connect pin 2 and pin 25
2 - Leave connector -Input
open
{Doc
}1
M2/
M4
Pro
toty
pe S
trai
n G
age
Boa
rd
A
22
Frid
ay, A
ugus
t 14,
199
8
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
Sen
se-L
ow 2
Vex
c_ou
t 2
Str
ain
2
Sen
se-H
igh
2+I
nput
2-I
nput
2
+12V
-12V
+9V
reg
-9V
reg
+12V
-12V
+12V
+12V
-12V
-12V
+-
U1B
LF34
7
567
411
R8
Str
ain
Gai
n A
djus
t 2
Strain Gage
Signal Conditioner
U5
1B31
1 2 3 4 8 9 10 11 12 13 141516171819202125262728
+ In
put
- In
put
Gai
n 1
Gai
n 2
Vou
t Unf
ilter
edIn
put O
ffset
Adj
1In
put O
ffset
Adj
2O
utpu
t Offs
et A
djB
andw
idth
Adj
. 1B
andw
idth
Adj
. 2V
out F
ilter
ed-
Vs
Com
m+
Vs
+ V
s R
egR
ef. O
utR
ef. I
nE
xc. A
dj
Hal
f-B
ridge
Com
plet
ion
Sen
se L
owS
ense
Hig
hV
exc
Out
R9
15k
C8
4.7u
F
R10
50k
1 3
2
+-
U1C
LF34
7
1098
411
+-
U1D
LF34
7
121314
411
Figure B-8: Page 2 of 2 of the strain gauge conditioning board.
APPENDIX B. ELECTRICAL SCHEMATICS 48
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
{Doc
}{R
evC
ode}
M2
BR
EA
KO
UT
BO
AR
D
A
11
Mon
day,
Mar
ch 0
6, 2
000
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
DA
C0+
DA
C1+
DA
C2+
DA
C3+
DA
C4+
DA
C5+
DA
C6+
DA
C7+
DA
C8+
DA
C9+
DA
C10
+D
AC
11+
DA
C12
+D
AC
13+
DA
C14
+D
AC
15+
AD
C+3
2A
DC
+33
AD
C+3
4A
DC
+35
AD
C+3
6A
DC
+37
AD
C+3
8A
DC
+39
AD
C+4
0A
DC
+41
AD
C+4
2A
DC
+43
AD
C+4
4A
DC
+45
AD
C+4
6A
DC
+47
DA
C0-
DA
C1-
DA
C2-
DA
C4-
DA
C3-
DA
C5-
DA
C6-
DA
C7-
DA
C8-
DA
C9-
DA
C10
-D
AC
11-
DA
C12
-D
AC
13-
DA
C14
-D
AC
15-
AD
C-3
2A
DC
-33
AD
C-3
4A
DC
-35
AD
C-3
6A
DC
-37
AD
C-3
8A
DC
-39
AD
C-4
0A
DC
-41
AD
C-4
2A
DC
-43
AD
C-4
4A
DC
-45
AD
C-4
6A
DC
-47
AD
C0+
AD
C0-
AD
C1+
AD
C1-
AD
C2+
AD
C2-
AD
C3+
AD
C3-
AD
C4+
AD
C4-
AD
C5-
AD
C5+
AD
C6+
AD
C6-
DA
C7+
AD
C7-
AD
C8+
AD
C8-
AD
C9-
AD
C9+
AD
C10
+A
DC
10-
AD
C11
+A
DC
11-
AD
C12
+A
DC
12-
AD
C13
-A
DC
13+
AD
C14
+A
DC
14-
AD
C15
+A
DC
15-
AD
C0+
AD
C21
+
AD
C11
+
AD
C13
-
AD
C22
+
AD
C12
+
AD
C14
-
AD
C23
+A
DC
24-
AD
C13
+
AD
C25
AD
C14
+
AD
C26
-
AD
C28
-
AD
C1-
AD
C1+
AD
C26
+A
DC
27-
AD
C2-
AD
C2+
AD
C17
+
AD
C27
+
AD
C3
AD
C3+
AD
C18
+A
DC
19-
AD
C28
+
AD
C4-
AD
C4+
AD
C8+
AD
C15
-
AD
C19
+A
DC
20-
AD
C24
+
AD
C5+
AD
C9+
AD
C16
-
AD
C20
+A
DC
21-
AD
C25
+
AD
C29
+
AD
C10
+
AD
C15
+
AD
C17
-
AD
C22
-
AD
C30
+
AD
C0-
AD
C6+
AD
C16
+
AD
C18
-
AD
C23
-
AD
C7+
AD
C29
-A
DC
30-
AD
C31
+A
DC
31-
AD
C16
+A
DC
16-
AD
C17
-A
DC
17+
AD
C18
+A
DC
18-
AD
C19
+A
DC
19-
AD
C20
+A
DC
20-
AD
C21
-A
DC
21+
AD
C22
+A
DC
22-
AD
C23
+A
DC
23-
DA
C14
-D
AC
13+
DA
C15
-D
AC
12+
DA
C12
-D
AC
13-
DA
C14
+D
AC
15+
GN
D
AD
C11
-A
DC
10-
AD
C12
-
AD
C6-
AD
C8-
AD
C7-
AD
C9
AD
C5-
AD
C24
+A
DC
24-
GN
DA
DC
25+
AD
C25
-
AD
C26
+A
DC
26-
GN
DA
DC
27+
AD
C27
-
AD
C28
+A
DC
28-
GN
DA
DC
29+
AD
C29
-
AD
C30
+A
DC
30-
GN
DA
DC
31+
AD
C31
-
AD
C+3
2A
DC
-32
GN
DA
DC
+33
AD
C-3
3
AD
C+3
4A
DC
-34
GN
D
AD
C-3
5A
DC
+35
AD
C+3
6A
DC
-36
GN
D
AD
C-3
7A
DC
+37
AD
C+3
8A
DC
-38
GN
D
AD
C-3
9A
DC
+39
AD
C+4
0A
DC
-40
GN
D
AD
C-4
1A
DC
+41
AD
C+4
2A
DC
-42
GN
D
AD
C-4
3A
DC
+43
AD
C+4
4A
DC
-44
GN
D
AD
C-4
5A
DC
+45
AD
C+4
6A
DC
-46
GN
D
AD
C-4
7A
DC
+47
DA
C1+
DA
C1-
GN
D
DA
C0+
DA
C0-
DA
C2+
DA
C2-
GN
D
DA
C3-
DA
C3+
DA
C4+
DA
C4-
GN
D
DA
C5-
DA
C5+
DA
C6+
DA
C6-
GN
D
DA
C7-
DA
C7+
DA
C8+
DA
C8-
GN
D
DA
C9-
DA
C9+
DA
C10
+D
AC
10-
GN
D
DA
C11
-D
AC
11+
JP1
HE
AD
ER
64
12
34
56
78
910
1112
1314
1516
1718
1920
2122
2324
2526
2728
2930
3132
3334
3536
3738
3940
4142
4344
4546
4748
4950
5152
5354
5556
5758
5960
6162
6364
12
34
56
78
910
1112
1314
1516
1718
1920
2122
2324
2526
2728
2930
3132
3334
3536
3738
3940
4142
4344
4546
4748
4950
5152
5354
5556
5758
5960
6162
6364 IN
-SH
IN1
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
IN-S
HIN
2
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
IN-T
HIG
H1
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
IN-T
HIG
H2
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
IN-B
OD
Y1
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
IN-B
OD
Y2
10P
IN_c
ople
ysty
le
1 2 3 4 5 6 7 8 9 10
A+
A-
GN
DA
V+
AV
-
B+
B-
GN
D2
BV
+B
V-
OU
T-X
4
3pin
_cop
ley_
styl
e
1 2 3
sig+
sig-
GN
D
OU
T-X
3
3pin
_cop
ley_
styl
e
1 2 3
sig+
sig-
GN
D
OU
T-X
2
3pin
_cop
ley_
styl
e
1 2 3
sig+
sig-
GN
D
OU
T-X
1
3pin
_cop
ley_
styl
e
1 2 3
sig+
sig-
GN
D
JP2
HE
AD
ER
64
12
34
56
78
910
1112
1314
1516
1718
1920
2122
2324
2526
2728
2930
3132
3334
3536
3738
3940
4142
4344
4546
4748
4950
5152
5354
5556
5758
5960
6162
6364
12
34
56
78
910
1112
1314
1516
1718
1920
2122
2324
2526
2728
2930
3132
3334
3536
3738
3940
4142
4344
4546
4748
4950
5152
5354
5556
5758
5960
6162
6364
IN-R
-FO
OT
1
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-R
-FO
OT
2
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-L
-FO
OT
1
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-L
-FO
OT
2
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
1
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
2
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
3
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
4
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
5
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
6
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
7
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
IN-X
8
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
OU
T-S
HIN
1
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
OU
T-S
HIN
2
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
OU
T-T
HIG
H1
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
OU
T-T
HIG
H2
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
OU
T-B
OD
Y1
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
OU
T-B
OD
Y2
5PIN
_cop
leys
tyle
1 2 3 4 5
A+
A-
GN
DB
+B
-
GN
D2
GN
D
Tes
t Poi
nt
GN
D1
GN
D0
Tes
t Poi
nt
GN
D3
GN
D1
Tes
t Poi
nt
Figure B-9: Page 1 of 1 of the analog I/O breakout board. This board is designed to interface withthe ANA070 and ANA064 from Digital Designs and Systems.
APPENDIX B. ELECTRICAL SCHEMATICS 49
5 5
4 4
3 3
2 2
1 1
DD
CC
BB
AA
Second power brick is either +5 or +12 single ended supply.
Power Board, February 2000
{Doc
}{R
evC
ode}
MIT
Leg
Lab
orat
ory
A
11
Tue
sday
, Feb
ruar
y 15
, 200
0
Titl
e
Siz
eD
ocum
ent N
umbe
rR
ev
Dat
e:S
heet
of
+12
-12
+12
-12
+12
+12
+12
+12
+12
-12
-12
-12
-12
+V
48+
48-
48-
48+
48-
48+
+V
+V
+12
-12
-12
-5-5+5
+5
+5
-5
5+
5-
DC/DC
CONVERTER
+-12V
PB
1 DF
C10
U48
D12
1 2
3 4 5
+IN
PU
T
-IN
PU
T
+12O
UT
CM
N
-12O
UT
PC
1
1 2
TP
4
plus
5
Tes
t Poi
nt
PC
510
uf
1 2
RE
G1
I434
06
13
2
IN+5
OU
T
GND
TP
5m
inus
5
Test Point
PC
6
10uf
12
RE
G2
I433
14
13
2
GN
D-5
OU
T
IN
J4 2pin
_ter
min
al_b
lock
_ed1
609
12
+VGN
D
J3
3pin
_ter
min
al_b
lock
_ed1
610
1 2 3
+V GN
D
-V
TP
2
min
us12
Tes
t Poi
nt
TP
1pl
us12
Tes
t Poi
nt
J2
3pin
_ter
min
al_b
lock
_ed1
610
1 2 3
+V GN
D
-V
U1
3pin
1 2 3
+5 out
-5 U3
3pin
1 2 3
+5 out
-5
U4
3pin
1 2 3
+5 out
-5U2
3pin
1 2 3
+5 out
-5
PC
210
uf
1 2
PC
310
uf
1 2
J5 2pin
_ter
min
al_b
lock
_ed1
609
12
+VGN
D
J1 48IN
_CO
NN1 2
+48
GN
D
TP
3
plus
5 o
r 12
Tes
t Poi
nt
PC
410
uf
12
DC/DC
CONVERTER
+5V
PB
2
DF
A6U
48S
5
12
345
+IN
PU
T
-IN
PU
T
+5O
UT
no c
onne
ctio
n
GN
D
J6
3pin
_ter
min
al_b
lock
_ed1
610
1 2 3
+V GN
D
-V J7
3pin
_ter
min
al_b
lock
_ed1
610
1 2 3
+V GN
D
-V
GN
D2
grou
nd
Tes
t Poi
nt
GN
D1
grou
nd0
Tes
t Poi
nt
Figure B-10: Page 1 of 1 of the M2 power distribution board. This board has two different populationoptions. One to power the DSP and one to power the Intersense IS-300 tracker.
Appendix C
Suppliers and Costs
This section contains a list of all the suppliers that were used in development of the robot.——————————————————————————–
Digital Designs and Systems3266 North Meridian StreetIndianapolis, IN 46208317-931-8190http://www.dideas.com/microcontrollers, a/d and d/a conversion, vision systems, ready built and custom designs
——————————————————————————–Stock Drive Products2101 Jericho Turnpike,Box 5416New Hyde Park, NY 11042-5416(516)328-3300Mechanical Parts
——————————————————————————–Berg499 Ocean AvenueBox 130East Rockaway, NY 11518(516)596-1700Mechanical Parts
——————————————————————————–NOVOTECHNIK237 CEDAR HILL STREETMARLBORO, MA 01752(508) 485-2244Potentiometers
——————————————————————————–SMALL PARTS INC13980 NW 58TH COURTPO BOX 4650MIAMI LAKES, FL 33014-3115(800) 220-4242Mechanical Parts
——————————————————————————–CENTURY SPRING COMPANY INCBOX 15287222 EAST 16TH STREET
50
APPENDIX C. SUPPLIERS AND COSTS 51
LOS ANGELES, CA 90015-0287(800)237-5225Springs
——————————————————————————–Newark Electronics41 Pleasant StreetMethuen, MA 01844(800) 463-9275Electronic Parts
——————————————————————————–REC ENGINEERING COMPANY(Bob)20 HOPKINS STREETWILMINGTON, MA 01887-2210(508)657-6517Machining
——————————————————————————–Eastern Tool(Joe)Somerville, MA(617)497-6703Machining
——————————————————————————–Transducer Techniques43178 Business Pk. Dr.Temecula CA 925901-800-344-3965Load cells.
——————————————————————————–Intersense, Inc.73 Second AvenueBurlington MA 01803(781)-270-0090Multi-axis inclinometer.
——————————————————————————–Digi-Key Corporation701 Brooks Ave. SouthThief River Falls, MN 56701-0677(800)344-4539Electronic Parts
——————————————————————————–Copley Controls Corporation410 University AvenueWestwood, MA 02090(617)329-8200FAX: (617)329-4055Amplifiers
——————————————————————————–Cooner Wire Company9265 OwensmouthChatsworth, CA 91311(818)882-8311Flexible Wire
APPENDIX C. SUPPLIERS AND COSTS 52
——————————————————————————–CFC, Inc.179 Bear Hill RoadWaltham, MA 02154-1001(617)890-1878FAX: (617)890-7098MODEM: (617)890-7193Circuit Board Fabrication
——————————————————————————–Advanced CircuitsColorado, http://www.4pcb.comCircuit Board Fabrication, online ordering
——————————————————————————–NSK -linear motion divisionElmhurst, IL 60126-1016ATTN: Sandy x2638 or Jackie x2677tel:1-800-255-4773fax:630-924-8197Ballscrews
——————————————————————————–Litton Poly-Scientific1213 North Main StreetBlacksburg, VA 24060(828)837-5115Attn: Barbara Smith x231fax (828)837-0846Brushless DC motors
——————————————————————————–McMaster CarrBOX 440NEW BRUNSWICK, NJ 08903-0440(908)329-3200Mechanical Parts
——————————————————————————–QUALITY COMPOSITES INC8385 S ALLEN STREET No.140SANDY, UT 84070(801)565-8003FAX: (801)565-8225Carbon Fiber Tubes
——————————————————————————–Sava Industries, Inc.4 N. Corporate Drive, P.O. Box 30Riverdale, NJ 07457-0030(201) 835-0882Steel Cable, crimps and crimp tools.
——————————————————————————–Tristar PlasticsShrewsbury, MA800.874.7827x3203(mike rudel)fax 508.845.1200
APPENDIX C. SUPPLIERS AND COSTS 53
Rulon bushings——————————————————————————–
W.S. Deans10875 Portal DriveLos Alamitos CA 90720714-828-6494 RobinWet noodle wire
——————————————————————————–US Sensor Corp1832 West Collins Ave.Orange, CA 92867714-639-1000Terry ext 104Thermal sensors for mounting amps to heat sinks
——————————————————————————–Tra-Con45 Wiggins Ave.Bedford, MA 017301.800.872.2261781.275.6363Glue for glueing thermal sensors to amp heat sink.
——————————————————————————–Tower HobbiesPO Box 9078Champaign, IL 61824-90781-800-637-6050Batteries and charger
——————————————————————————–
C.1 Costs
This section contains the costs associated with the robot M2. Costs and quantities given do notinclude spares. The general lab policy is to order 50 percent spares on everything. I think thatanyone else developing a similar robot should do the same. See Table C.1 for details.
APPENDIX C. SUPPLIERS AND COSTS 54
Component QTY Each US $ Total US $Actuator, Amp 12 1500 18,000
Custom Computer 1 10,000 10,000CFRP frame 8 300 2500Machining 1 30,000 30,000Vestibular 1 10,000 10,000Misc Parts 200 20 4000
PCB’s 10 1,000 10,000Load Cells 8 450 3600TOTALS 88,100
Table C.1: Basic robot budget. See Robinson et al.((17)) for more detail on the actuator. Thisbudget does not include prototyping or development costs. Part quantities and individual costs arenot necessarily meant to imply identical parts but rather to give an average price for all units of acertain type.