ECE 43407340 Exam

1
ECE 4340/7340Exam 2 Review

• Winter 2005

2
Sensing and Perception

• CMUcam and image representation (RGB, YUV)
• Percept logical sensors
• Logical redundancy vs. physical redundancy
• Combining sensory signals
• Sensor fission
• Sensor fashion
• Sensor fusion

3
Sensor Fission
Sensor Fashion
Sensor Fusion
4
Sensory Uncertainty (4.2-4.3)

• Gaussian distribution of input data
• Uncertainty propagation to output
• Line extraction from noisy range data

5
Angle Histograms using Range (4.3)
6
Architectures

• Subsumption Brooks
• One behavior takes precedence at a time
• AuRA Arkin (hybrid)
• Potential fields for navigation
• Piecewise linear paths from landmark to landmark
• Be prepared to design a potential field approach
  for a designated problem (e.g., docking)

7
Using Schemas for Robot Behaviors

• Perceptual schema Motor schema
• Behavior NOT a function or an event

motor actions
sensor input
Perceptual Schema
Motor Schema
percept gain
8
Wander for color
Move to color
Wander for light
Move to light
Release color
Include inputs to behaviors!
9
Mataric

• Topological mapping, planning navigation using
  the subsumption architecture
• Range sensors, compass Sensor perceptual zones
• What constitutes a landmark?
• How are landmarks recognized?
• Map representation
• Graph where each node is a landmark
• Zero distance between nodes
• How was planning accomplished?

10
Other Topological Map Representations
node
Connectivity (arch)
11
Chapter 5

• Probabilistic map-based localization (5.6)
• Action update based on wheel encoders
• Perception update based on sensors in new
  location
• Dervish example

12
Kuipers

• Layers
• Geometric level
• Topological level
• Sensorimotor Control level
• Distinctive places
• a local maximum found by a hill-climbing
  strategy

13
Levitt and Lawton

• Triangular-shaped regions formed by landmarks
• Topological planning navigation from region to
  region
• How was planning accomplished?

14
Chapter 6

• Configuration space for mobile robots
• Representations
• Visibility graph
• Voronoi diagram
• Cell decomposition (e.g., grid cell)
• Path planning / search algorithms
• NF1 or grassfire
• Graph search Breadth first, Depth first, Greedy,
  A
• Obstacle avoidance
• Potential field,
• Bug1, Bug2,
• Vector field histogram

Be prepared for a search problem for planning
15
A search for path planningFor search, distance
actual distance to node estimated distance
16
Balch and Arkin

• Robot formations as motor schemas
• Diamond, wedge, line, follow the leader
• Control referencing
• Leader, neighbor, unit
• Zones
• Ballistic, controlled, deadzone
• Results

17
Parker - ALLIANCE

• Multi-robot distributed coordination
• Impatience
• Acquiescence
• Extension of Subsumption
• Behavior sets are switched out to give each robot
  its role
• Each robot broadcasts its activity
• Results

18
Murphy and Lisetti

• Multi-robot distributed coordination via emotions
• Multi-agent control for interdependent tasks
• Cyclic dependency
• Emotional states change each robots behavior
• Frustrated
• Concerned
• Confident
• Happy
• Why do we insist on using biological models for
  robot behavior when it is not necessary?