Probabilistic Robotics and Models of Gaze Control

José Ignacio Núñez Varela

Seminario del Posgrado de Ingeniería Eléctrica 22 noviembre 2013

• Introduction to probabilistic robotics

• Introduction to the problem of gaze control

• Candidate models of gaze control

• Experiments and conclusions

• Research projects

Outline:

Part I:Probabilistic Robotics

This part is mainly based on Chapters 1 and 2 of the book: Thurn et al. Probabilistic Robotics, MIT Press, 2005.

Basic Model: Robotic systems are situated in the real world,

perceive information on their environment through sensors, and manipulate through physical forces.

Picture credits: http://www.3ders.org/images/PrimeSense_apple-3d-sensor-1.jpg

http://cdn.shopify.com/s/files/1/0130/8982/products/midi-cpu-large_1024x1024.jpg

Sensing Planning

Acting

Asimo © Honda

Intelligent robotics

Learning

Reasoning

Decision-making

Planning

Understanding

Common sense

PR2 © Willow Garage

Robots have to be able to accomodate the enormous uncertainty that exists in the physical world.

Imagen: http://www.grumpygratefulmom.com/wp-content/uploads/2011/11/messy-kitchen-1024x769.jpg

What factors contribute to the robot's uncertainty?

Imagen: http://www.grumpygratefulmom.com/wp-content/uploads/2011/11/messy-kitchen-1024x769.jpg

Robot Environments:

Inherently unpredictableUncertainty is high for robots operating in the

proximity of people

Robot Environments:

Well structured environment>> uncertainty

Robot Sensors:

Sensors are limited in what they can perceiveE.g., physical limitations affect range and resolutionSensors are subject to noiseSensors can break

Robot Actuators:

Motors are, at some extent, unpredictableControl noise, wear-and-tear, mechanical failure

Robot's Internal Models (software):

All internal models of the world are approximateModel errors have often being ignored

Algorithmic Approximations (software):

Robots are real time systems, thus limiting the amount of computation being carried out

Algorithms need to be approximated

Uncertainty

Robots are forced to act even though it doesn't have sufficient information to make decisions with absolute certainty

As robots are now moving into the open world, uncertainty becomes a major issue

“Managing uncertainty is possibly the most important

step towards robust real-world robot systems”

- Thurn, Burgard and Fox

Probabilistic Robotics

Key idea: Represent uncertainty explicitly using the calculus of probability theory

Instead of relying on a single “best guess”, probabilistic algorithms represent information by probability distributions over a whole space of guesses

Mobile Robot Localization:

The problem of estimating a robot's coordinates relative to an external reference frame (the robot is given a map)

A probability density function over the space of all locations represents the robot's belief

This belief is updated using the robot's sensors

Mobile Robot Localization:

Major Paradigms

Model-basedrobotics

Probabilisticrobotics

Behaviour-based

robotics

Mid-1970s Mid-1990s Mid-1980s

Model-based vs. Probabilistic Robotics:

Model-based robotics require accurate models of the robot, environment, etc.

Probabilistic robotics have weaker requirements on this accuracy

Behaviour-based vs. Probabilistic Robotics:

Behaviour-based robotics require accurate sensorsProbabilistic robotics have weaker requirements on this

accuracy

Part II:Gaze Control

Imagen: http://www.grumpygratefulmom.com/wp-content/uploads/2011/11/messy-kitchen-1024x769.jpg

Biological perspective

Gaze Control

Machine perspective

© Jason Babcock © icub.org

Why study gaze control?

© cellfield.ca

Foveal Vision

© Michael Land

Eye Movements

Saccades

• Rapid jump-like movements (900°/sec)• Ballistic (trajectory cannot change)• Stereotyped (follow the same pattern)• Voluntary and involuntary

• Aim: Shift the fovea to obtain high resolution samples

Saccade Sequence

We perform hundreds or even thousands of saccades every day!

How does the brain decide where to

fixate next?

© Ilya Repin

Active Vision

© Yarbus

Task and context determine where to

fixate next

Vision and Action

© Mary Hayhoe

Uncertainty Reduction

Engineering science goal

What mechanisms a rational decision maker could employ to select a gaze location optimally, or near optimally, given limited information and limited computation time during the performance of a task?

Human behavioural goal

How humans select the next gaze location?

Gaze Control Processes

iCub Humanoid Robot

© icub.org

Two problems

where to look gaze allocation

Pick & Place Task

Models of Gaze Control

• Based on uncertainty reduction (Uncertainty)

• Based on rewards and uncertainty (Rew+Unc)

• Based on rewards, uncertainty and gain (Rew+Unc+Gain)

“What would happen if I look at entity ei?”

One-step look ahead gaze control

Uncertainty Reduction

“How much uncertainty is reduced if I look at entity ei?”

X

Reward and Uncertainty

“How much value am I expected to get after looking

at entity ei?”

X

Reward, Uncertainty & Gain

“Which motor system would get more benefit if gaze is

allocated to it?”

X

Experiments

We characterise how task performance varies in terms of three environmental parameters:

•Reach/grasp sensitivity

•Observation noise

•Field of view

Also compared against Random and Round Robin gaze strategies

Reach/Grasp Sensitivity

Observation Noise

Field of View

Conclusions

• Gaze control models that incorporate rewards and uncertainty seem to perform better

• The Rew+Unc+Gain scheme has, in general, the best overall performance

• An active visual search process should be integrated into the Rew+Unc+Gain strategy

• The Rew+Unc+Gain and Rew+Unc gaze schemes were able to reproduce behavioural human data

Coordinación de Módulos de Control Guiados Visualmente en un Marco de

Toma de Decisiones para Robots Humanoides

Cesar A. Puente Montejano (responsable)Juan C. Cuevas TelloJosé I. Núñez VarelaOmar Vital OchoaFrancisco E. Martínez PérezOmar Rodríguez GonzálezRogelio Castillo MorquechoOctavio Rentería Vidales

Imagen: http://www.informatik.uni-hamburg.de/WTM/pictures/topics/robot_world_interaction.jpg

Imagen: http://www.unitec.ac.nz/advance/wp-content/uploads/2012/11/Robots4.jpg

http://www.pacman-project.euContact information: Jeremy L. Wyatt

Procesamiento de Señales Biomédicas

Carlos Soubervielle MontalvoOmar Vital OchoaJuan C. Cuevas TelloJosé I. Núñez Varela

Imagen: http://emotiv.com/upload/iblock/6ac/header_set.gif

© IPN

Thank You!!

E-mail: jose.nunez@uaslp.mxWebsite: http://ciep.ing.uaslp.mx/jnunez

© Botodesigns / Chen Reichert

Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29Slide 30Slide 31Slide 32Slide 33Slide 34Slide 35Slide 36Slide 37Slide 38Slide 39Slide 40Slide 41Slide 42Slide 43Slide 44Slide 45Slide 46Slide 47Slide 48Slide 49Slide 50Slide 51Slide 52Slide 53Slide 54Slide 55Slide 56Slide 57Slide 58Slide 59Slide 60Slide 61Slide 62Slide 63Slide 64Slide 65