Introduction to Autonomous Mobile Robots

Prof. Yan Meng

Department of Electrical and Computer Engineering

Stevens Institute of Technology
Course Logistics

z Instructor: Yan Meng

¾ Office: Burchard 411
¾ Phone: 201-216-5496
¾ Email: [email protected]
¾ Office hour: Tuesday 3:00pm-5:00pm
¾ Course website:
http://www.ece.stevens-
tech.edu/~ymeng/courses/CPE521/CPE521A.htm
z Homework
¾ Homework will be due one week later after it is assigned
¾ Problem solutions will be posted on-line – LATE HOMEWORK WILL
NOT BE ACCEPTED AFTER THE SOLUTION IS POSTED
z Grading
¾ Homework 20% Midterm 20% Final 30% Project 30%
Course Syllabus

z Required Textbook:
¾ Roland Siegwart and Ilah Nourbakhsh, Introduction to Autonomous Mobile
Robots, MIT Press, April 2004, ISBN# 0-262-19502-X.
¾ Textbook website: http://autonomousmobilerobots.epfl.ch/
¾ Some reading materials and hands out will be distributed in class.
z Recommended readings:
¾ George A. Bekey, Autonomous Robots – From Biological Inspiration to
Implementation and Control, MIT Press, 2005. ISBN 0-262-02578-7.
¾ Robin Murphy, An Introduction to AI Robotics, MIT Press, November 2000.
ISBN 0-262-13383-0.
¾ Stefano Nolfi and Dario Floreano, Evolutionary Robotics: The Biology,
Intelligence, and Technology of Self-Organizing Machines, MIT Press,
2000, ISBN 0-262-14070-5.
¾ Thomas Braunl, Embedded Robotics: Mobile Robot Design and
Applications with Embedded Systems, Springer-Verlag Berlin Heidelberg
New York, ISBN 3-540-03436-6.
Some Robotics Links

z http://www.ifi.unizh.ch/groups/ailab/links/robotic.html#companies
z http://www.cooper.edu/~mar/robotics_links.htm

z http://www.roboticsonline.com/links/

z http://www.ieee-ras.org/

z http://www.euronet.nl/users/ragman/link_64.html
 1
Applications of Mobile Robots

Indoor Outdoor
Structured Environments Unstructured Environments
 1
Autonomous Mobile Robots

z The three key questions in Mobile Robotics

¾ Where am I ?
¾ Where am I going ?
¾ How do I get there ?

z To answer these questions the robot has to

¾ have a model of the environment (given or autonomously built)
¾ perceive and analyze the environment
¾ find its position within the environment
¾ plan and execute the movement
z Basic tasks: deal with Locomotion and Navigation (Perception,
Localization, Planning and motion generation)
 1
Control of Mobile Robots

Knowledge, Mission
Data Base Commands ¾ Most functions for
save navigation are
’local’ not involving
Localization Cognition localization nor
"Position" Path Planning
Map Building Global Map
global cognition
Environment Model Path
Local Map ¾ Localization and
local Information Path global path planning
Extraction Execution Î slower update

Motion Control
rate, only when
Perception

Raw data Actuator Commands

needed

Sensing Acting ¾ This approach is

pretty similar to what
human beings do.
Real World
Environment
 1
Control Architectures / Strategies

z Control Loop
¾ dynamically changing
¾ no compact model available
¾ many sources of uncertainty

Localization "Position" Cognition

Global Map

Environment Model Path

Local Map

Perception Real World Motion Control

Environment
 1
Two Approaches

z Classical AI
(model based navigation)
¾ complete modeling
¾ function based
¾ horizontal
decomposition
z New AI
(behavior based navigation)
¾ sparse or no modeling
¾ behavior based
¾ vertical decomposition
¾ bottom up
z Possible Solution
¾ Combine Approaches
 Environment Representation and Modeling: 1
The Key for Autonomous Navigation

z Environment Representation
¾ Continuos Metric -> x,y,θ
¾ Discrete Metric -> metric grid
¾ Discrete Topological -> topological grid
z Environment Modeling
¾ Raw sensor data, e.g. laser range data, grayscale images
o large volume of data, low distinctiveness
o makes use of all acquired information
¾ Low level features, e.g. line other geometric features
o medium volume of data, average distinctiveness
o filters out the useful information, still ambiguities
¾ High level features, e.g. doors, a car, the Eiffel tower
o low volume of data, high distinctiveness
o filters out the useful information, few/no ambiguities, not enough information
 Environment Representation and Modeling: How we do it! 1

z Odometry z Modified z Feature-based

Environments Navigation

Corridor
crossing
Elevator door

Courtesy K. Arras
Entrance
How to find a treasure
Landing at night
Eiffel Tower
¾ not applicable ¾ expensive, ¾ still a challenge for
inflexible artificial systems
 1
Environment Representation: The Map Categories

z Recognizable Locations z Topological Maps

Courtesy K. Arras
z Metric Topological Maps z Fully Metric Maps (continuos or
discrete)
 1
Methods for Navigation: Approaches with Limitations

z Incrementally z Modifying the environments

(dead reckoning) (artificial landmarks / beacons)

Courtesy K. Arras
Inductive or optical tracks (AGV)

Odometric or initial
sensors (gyro) Reflectors or bar codes

¾ not applicable
¾ expensive, inflexible
 1
Methods for Localization: The Quantitative Metric Approach

1. A priori Map: Graph, metric 3. Matching:

Find correspondence

Courtesy K. Arras
of features

2. Feature Extraction (e.g. line segments) 4. Position Estimation:

e.g. Kalman filter, Markov

z representation of uncertainties
z optimal weighting acc. to a priori statistics
 1
Methods for Localization: The Quantitative Topological Approach

1. A priori Map: Graph 3. Library of driving behaviors

locally unique e.g. wall or midline following, blind step,
points enter door, application specific
behaviors
edges Example: Video-based navigation with
natural landmarks

2. Method for determining

the local uniqueness

e.g. striking changes on raw data level

or highly distinctive features Courtesy of [Lanser et al. 1996]
 1
Map Building: How to Establish a Map

1. By Hand 3. Basic Requirements of a Map:

¾ a way to incorporate newly sensed
information into the existing world
model
¾ information and procedures for
estimating the robot’s position
¾ information to do path planning and
other navigation task (e.g. obstacle
2. Automatically: Map Building
avoidance)

Courtesy K. Arras
predictability
The robot learns its environment
z Measure of Quality of a map
¾ topological correctness
Motivation:
¾ metrical correctness
- by hand: hard and costly
- dynamically changing environment z But: Most environments are a mixture of
predictable and unpredictable features
- different look due to different perception
→ hybrid approach
model-based vs. behaviour-based
 1
Map Building: The Problems

1. Map Maintaining: Keeping track of 2. Representation and

changes in the environment Reduction of Uncertainty

Courtesy K. Arras
e.g. disappearing position of robot -> position of wall
cupboard

position of wall -> position of robot

- e.g. measure of belief of each z probability densities for feature positions

environment feature z additional exploration strategies
 1
Map Building: Exploration and Graph Construction

1. Exploration 2. Graph Construction

Courtesy K. Arras
Where to put the nodes?

z Topology-based: at distinctive locations

- provides correct topology
- must recognize already visited location
z Metric-based: where features disappear or
- backtracking for unexplored openings get visible