Share Course Ware
Engineering > Aeronautics & Astronautics > Cognitive Robotics
 Cognitive Robotics  posted by  member150_php   on 2/22/2009  Add Courseware to favorites Add To Favorites  
Further Reading
More Options

Williams, Brian, 16.412J Cognitive Robotics, Spring 2005. (Massachusetts Institute of Technology: MIT OpenCourseWare), (Accessed 07 Jul, 2010). License: Creative Commons BY-NC-SA

Cognitive Robotics

Spring 2005

Image of the Martian surface, taken by the Mars Rover.

This image of the Martian surface was taken by NASA's Mars Exploration Rover Spirit on July 16, 2004. (Image courtesy of NASA/JPL.)

Course Highlights

This course features a full set of lecture notes and student projects from the term.

Course Description

Cognitive robotics addresses the emerging field of autonomous systems possessing artificial reasoning skills. Successfully-applied algorithms and autonomy models form the basis for study, and provide students an opportunity to design such a system as part of their class project. Theory and application are linked through discussion of real systems such as the Mars Exploration Rover.

Technical Requirements

Special software is required to use some of the files in this course: .wcsp, .zip.


Course Staff

Prof. Brian Williams


Modern and future robots will have enough computational horse power to be able to reason continuously and deeply about themselves and their environments. In this advanced graduate course we will study the models and algorithms underlying recent robotic successes, ranging from the Mars Exploration Rover and similar remote systems, to the Nursebot, Museum Tourguide, and similar human-interaction systems. We will discuss the theory that underlies these algorithms and how they are implemented on real systems.

This year potential course projects include the creation of a fully autonomous Mars rover, using NASA's Mars Rover simulator, and daily mission scenarios provided by the JPL Mars Exploration Rover science team.

Algorithms and paradigms for creating a wide range of cognitive systems that act intelligently and robustly, by reasoning extensively from models of themselves and their world. Examples include a wide range of embedded and robotic systems, including autonomous Mars explorers, cooperative vehicles, and human interaction systems. Topics include deduction and search in real-time; temporal, decision-theoretic and contingency planning; dynamic execution and re-planning; reasoning about hidden state and failure; reasoning under uncertainty, path planning, mapping and localization, and cooperative and distributed robotics.


6.041 and either 16.410, 16.413, 6.034, or 6.825. Programming proficiency assumed.


Course schedule.

Lec #




Introduction to Cognitive Robotics
Learning Objectives, Remote Explorers, Model-based Programming

Students fill out sign-up sheet and review candidate lectures

Robots that Deftly Navigate


Kinodynamic and Randomized Path Planning
Review of Configuration Spaces, Visibility Graphs, Voronoi Diagrams, Potential Fields, and Cell Decomposition
Kino-dynamic Planning, Planning with Moving Obstacles, Probabilistic Roadmaps (PRMs), Rapidly Exploring Random Trees (RRTs)



Introduction to Simultaneous Localization and Mapping (SLAM) (Guest: Paul Robertson)
Localization, SLAM, Kalman Filter, Large Scale SLAM

Problem set 1 due
Student survey responses discussed


Vision Based SLAM (Guest: Paul Robertson)
Topological Maps, Hidden Markov Models (HMM), SIFT, Vision-based Localization


Deducing State and Diagnosing Failure


Model-based Diagnosis and Mode Estimation

Consistency-based Diagnosis: Candidates, Conflicts, Diagnoses, and Kernel Diagnoses
Conflict Extraction and Candidate Generation, Mode Estimation and Probabilistic Diagnosis, Active Probing



Solving Optimal CSPs through Conflict-Learning

Optimal Constraint Satisfaction Problems, Constraint-based A*, Conflict-directed A*, Conflict Extraction

Problem set 2 due
Benchmark examples reviewed

Reasoning About Soft Constraints


Soft Constraint Satisfaction Problems (SCSPs) (Guest: Martin Sachenbacher)
Valued Constraint Satisfaction Problems (VCSPs), Branch-and-bound Search for Soft Constraints, Variable Elimination for Soft Constraints, Tree Decomposition, Dynamic Programming



Solving CSPs and SCSPS via Decomposition and Abstraction (Guest: Martin Sachenbacher)
Reduced Ordered Binary Decision Diagrams (ROBDDs), Representing and Manipulating Soft Constraints using Algebraic Decision Diagrams (ADDs)


Planning Complex Missions


Mission-level Task Planning (Guest: Robert Tappan Morris)
Partial Order Planning, Constraint-based Interval Planning, and Simple Temporal Networks (STNs)



Dynamic Plan Execution Under Uncertainty
STNS, Dispatchable Networks and Dispatching Execution, STNUs, Strong and Dynamic Controllability



Mixed Human Robotic Exploration (Guest: Jeff Hoffman)

Problem set 3 due

Robots that Plan on the Fly


Hidden State and Model-based Reactive Planning
Universal Planning, Structure Decomposition for Model-based Reactive Planning (MRP), Binary Decision Diagrams, Symbolic MRP



Continuous, Incremental Path Planning and Exploration
Single Source Shortest Path, D*, LRTA*

Advanced lecture schedule assigned


Planning with POMDPs (Student Presenters: Brian Bairstow, Tony Jimenez, and Larry Bush)
An Introduction to the Fundamentals of POMDPs, State of the Art in POMDP Research, A Pedagogical Explanation of the Respective Algorithm



Model-based, Multi-Agent Reasoning in Texas Hold'em Poker (Student Presenters: Brian Edward Mihok and Michael Terry)
Leading Techniques in Games Reasoning, Emphasis on Uncertainty Techniques
Hidden Markov Models and Bayesian Inference, Neural Networks



Cognitive Game Theory (Student Presenters: Justin Fox, Jeremie Pouly, and Jennifer Novosad)
Alpha-Beta and its Extensions
An Evolutionary Algorithm Applied to Chess
Inductive Adversary Modeler

Problem set 4 due


Mode Estimation for Hybrid Discrete/Continuous Systems (Student Presenters: Lars Blackmore)
Trajectory Tracking for Constraint-based HMMs, Gaussian Filtering for Hybrid HMMs (K-Best and Rao-Blackwell Particle Filtering)



Particle Filters and their Applications (Student Presenters: Kaijen Hsiao, Jason Miller, and Henry Lefebvre de Plinval-Salgues)
Particle Filters in SLAM in Fault Diagnosis



Hello Computer? (Student Presenters: Shuonan Dong, Shen Qu, and Thomas Coffee)
SharedPlan, Plan Recognition, and COLLAGEN



Advanced Topics in Bayesian Networks (Student Presenters: Tom Temple, Ethan Howe, and James Lenfestey)
Dynamic Bayes Networks, Exact Inference, Approximate Inference (PF), Learning, Probabilistic Relational Models, Parameter/Structure Estimation


Sensing and Manipulating at the Cognitive Level


Visual Interpretation using Probabilistic Grammars (Guest: Paul Robertson)
Statistical Parsing, Image Segmentation, Monte Carlo Methods, Language Learning



Safe Execution of Bipedal Walking Tasks (Guest: Andreas Hoffman)
Motivation and Requirements, Bipedal Balance Control Strategies, Common Control Approaches (and their Failings), Task-level Control using Model-based Executives, Whole-body Control


Human - Robot Interaction


Working with and Learning from Humans as Partners (Guest: Cynthia Breazeal)
Multi-modal Communication, Human-robot Teamwork, Socially Guided Learning



Nursebot: Dialogue as a Decision Making Process (Guest: Nick Roy)
Model-based Dialog Management, Hierarchical Planning under Uncertainty, Reinforcement Learning for Human Interaction



Project Demonstrations



Project Demonstrations (cont.)   Tell A Friend