CpE 521A

A Introduction to Autonomous Mobile Robots

Lecture 7: Localization and Map Building

Part 1

Yan Meng
Department of Electrical and Computer Engineering
Stevens Institute of Technology
Localization and Map Building
• Noise and aliasing; odometric position estimation
• To localize or not to localize
• Belief representation
• Map representation

Localization "Position" Cognition

Global Map

Environment Model Path

Local Map

Perception Real World Motion Control

Environment
Localization, Where am I?

• Odometry, Dead Reckoning

• Localization base on external sensors,

Perception
beacons or landmarks
• Probabilistic Map Based Localization
Challenges of Localization
• GPS may be the answer?
• Knowing the absolute position (e.g. GPS) is not sufficient
¾ GPS can not function indoors or in obstructed areas
• Localization in human-scale in relation with environment
• Planning in the Cognition step requires more than only position as input
¾ It may need to acquire or build a map
• Localization actually means building a map, then identifying the robot’s
position relative to the map
• Perception and motion play important roles
¾ Sensor noise
¾ Sensor aliasing
¾ Effector noise
¾ The inaccuracy and incompleteness of sensors and effectors pose the
difficult challenges to localization
Sensor Noise
• Sensor noise induces a limitation on the consistency of sensor readings
in the same environmental state

• Sensor noise is mainly influenced by environment

e.g. surface, illumination …

• or by the measurement principle itself

e.g. interference between ultrasonic sensors

• Sensor noise drastically reduces the useful information of sensor

readings. The solution is:
¾ to take multiple reading into account
¾ employ temporal and/or multi-sensor fusion
Sensor Aliasing
• Unlike human sensing system, in robots, non-uniqueness of sensors
readings is the norm
¾ Example: sonar or laser rangefinder

• Even with multiple sensors, there is a many-to-one mapping from

environmental states to robot’s perceptual inputs

• Therefore the amount of information perceived by the sensors is

generally insufficient to identify the robot’s position from a single
reading
¾ Robot’s localization is usually based on a series of readings
¾ Sufficient information is recovered by the robot over time
Effector Noise: Odometry and Dead Reckoning
• Odometry and dead reckoning: position update is based on
proprioceptive sensors
¾ Odometry: wheel sensors only
¾ Dead reckoning: also heading sensors

• The movement of the robot, sensed with wheel encoders and/or

heading sensors is integrated to the position.
¾ Pros: Straight forward, easy
¾ Cons: Errors are integrated -> unbound

• Using additional heading sensors (e.g. gyroscope) might help to reduce

the cumulated errors, but the main problems remain the same.
Odometry: Error Sources
deterministic non-deterministic
(systematic) (non-systematic)

¾ Deterministic errors can be eliminated by proper calibration of the system.

¾ Non-deterministic errors have to be described by error models and will always
leading to uncertain position estimate.
• Major Error Sources:
¾ Limited resolution during integration (time increments, measurement resolution
…)
¾ Misalignment of the wheels (deterministic)
¾ Unequal wheel diameter (deterministic)
¾ Variation in the contact point of the wheel
¾ Unequal floor contact (slipping, not planar …)
¾…
Odometry: Classification of Integration Errors
• Range error: integrated path length (distance) of the robots movement
¾ sum of the wheel movements
• Turn error: similar to range error, but for turns
¾ difference of the wheel motions
• Drift error: difference in the error of the wheels leads to an error in the
robots angular orientation
Over long periods of time, turn and drift errors
far outweigh range errors!
¾ Consider moving forward on a straight line along the x axis. The error
in the y-position introduced by a move of d meters will have a component
of dsinΔθ, which can be quite large as the angular error Δθ grows.
Odometry: The Differential Drive Robot (1)
• Position can be estimated starting from a know position p by
integrating the movement
⎡ ⎤ x ⎡ Δx ⎤
p = ⎢ y⎥ ′ ⎢ ⎥
p = p + ⎢ Δy ⎥
⎢ ⎥
⎢⎣ θ ⎥⎦ ⎢⎣Δθ⎥⎦
Odometry: The Differential Drive Robot (2)
• Kinematics
Odometry: The Differential Drive Robot (3)
• Error model
Odometry: Growth of Pose Uncertainty for Straight Line Movement

• Note: Errors perpendicular to the direction of movement are growing much faster!
Odometry: Growth of Pose uncertainty for Movement on a Circle
• Note: Errors ellipse in does not remain perpendicular to the direction of movement!
Odometry: Calibration of Errors I (Borenstein [5])
• The unidirectional square path experiment

• BILD 1 Borenstein
Odometry: Calibration of Errors II (Borenstein [5])
• The bi-directional square path experiment

• BILD 2/3 Borenstein

To localize or not?
• How to navigate between A and B
¾ navigation without hitting obstacles
¾ detection of goal location
• Possible by always following the left wall
¾ However, how to detect that the goal is reached
Behavior Based Navigation

Since sensors and effectors are noisy and information-limited, one may
want to design sets of behaviors instead of creating a geometric map
for localization.

Assume there exists a procedural solution to the particular navigation

problem at hand.
Behavior Based Navigation
• Advantage
¾ May be implemented very quickly for a single environment with a small
number of goal positions

• Disadvantages
¾ Does not directly scale to other environments or to larger environments
¾ The underlying procedures must be carefully designed to produce the
desired behavior ( time-consuming and environmental-dependent)
¾ The fusion and rapid switching between multiple behaviors can negate
the fine-tuning procedure, and the addition of new behavior forces the
designer to retune all of the existing behaviors again
Model Based Navigation

Explicitly attempts to localize by collecting sensor data, then updating some

belief about its position with respect to a map of the environment
Model Based Navigation
• Advantage
¾ The explicit, map-based concept of position makes the system’s belief
about position transparently available to the human operators
¾ The existence of the map itself represents a medium for communication
between human and robot
¾ The map, if created by the robot, can be used by humans as well,
achieving two uses

• Disadvantage
¾ Require more up-frond development effort to create a navigating robot
¾ If the model diverges from reality (i.e., if the map is wrong), then the
robot’s behavior may be undesirable, even if the raw sensor values of the
robot are only transiently incorrect
Belief Representation
• a) Continuous map
with single hypothesis

• b) Continuous map
with multiple hypothesis

• d) Discretized map
with probability distribution

• d) Discretized topological
map with probability
distribution
Belief Representation: Characteristics

• Continuous • Discrete
¾ Precision bound by sensor ¾ Precision bound by
data resolution of discretisation
¾ Typically single hypothesis ¾ Typically multiple hypothesis
pose estimate pose estimate
¾ Lost when diverging (for ¾ Never lost (when diverges
single hypothesis) converges to another cell)
¾ Compact representation and ¾ Important memory and
typically reasonable in processing power needed.
processing power. (not the case for topological
maps)
 Single-hypothesis Belief – Continuous Maps

Real map with walls, doors, and furniture Line-based map

Single-hypothesis Belief – Grid and Topological Map

Topological map using

Occupancy grid-based map
line features and doors
Single Hypothesis Belief
• Advantage:
¾ No position ambiguity
¾ Make the decision-making much easier

• Disadvantage:
¾ Always generate a single hypothesis for position update is challenging
due to the effector and sensor noise
Multiple-hypothesis Belief
• The robot tracks not just a single possible position but a possibly
infinite set of positions
• One way to represent the set of possible robot positions is to use
multiple Gaussian probability density functions
• Advantages
¾ Maintain a sense of position while explicitly annotating the robot’s
uncertainty about its own position
¾ Enable robots with limited sensory information to navigate robustly
• Disadvantages
¾ Make the decision-making more difficult
¾ Some of the robot’s possible positions imply a motion trajectory that is
inconsistent with some of its other possible positions
¾ Computational expensive
Grid-base Representation - Multi Hypothesis
• Grid size around 20 cm2.

Courtesy of W. Burgard
Map Representation
1. Map precision vs. application

2. Features precision vs. map precision

3. Precision vs. computational complexity

• Continuous Representation

• Decomposition (Discretization)
Representation of the Environment
• Environment Representation
¾ Continuos Metric → x,y,θ
¾ Discrete Metric → metric grid
¾ Discrete Topological → topological grid
• Environment Modeling
¾ Raw sensor data, e.g. laser range data, grayscale images
o large volume of data, low distinctiveness on the level of individual values
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
Map Representation: Continuous Line-Based
a) Architecture map
b) Representation with set of infinite lines
Map Representation: Decomposition (1)
• Exact cell decomposition
¾ Exact decomposition is not always feasible in real-world
Map Representation: Decomposition (2)
• Fixed cell decomposition
¾ Obstacle-filled or free area
¾ Narrow passages disappear
Map Representation: Decomposition (3)
• Adaptive cell decomposition
Map Representation: Decomposition (4)
• Occupancy grid – with very small cells
¾ Most common map representation technique currently utilized
¾ Memory size may become untenable with large size of environment, not
compatible with closed-world assumption

Courtesy of S. Thrun
Map Representation: Decomposition (5)

• Topological Decomposition
Map Representation: Decomposition (6)
• Topological Decomposition

node

Connectivity
(arcs)
Map Representation: Decomposition (7)
• Topological Decomposition
State-of-the-Art: Current Challenges in Map
Representation
• Real world is dynamic
¾ Differentiate permanent obstacles (e.g., walls, doorways, etc.) and
transient obstacles (e.g., humans, shipping packages, etc.)
¾ Perception is still a major challenge (error prone, extraction of useful
information difficult)
• Traversal of open space
¾ Traditional range sensors are difficult for wide-open spaces, such as
parking lots, fields of grass, and indoor atriums, because of their relative
sparseness
• How to build up topology (boundaries of nodes) in wide-open area?
¾ GPS may be one solution
• Sensor fusion
¾ A variety of sensor types can have their data correlated appropriately,
obtain the perceptions well beyond that of any individual one
• More …