Lecture 8 - Emergence & Localization

Question 1

Synonym for arbitration-based behavior coordination (action selection)

Answer 1

Competitive behavior coordination

Question 2

Behavior arbitration
Competitive behavior coordination
(2) Different hierarchies
+ in which context are they used

Answer 2

1. Fixed priority hierarchy (pre-assigned priorities, used in subsumption architecture)
2. Dynamic priority hierarchy (priorities can change over time, used in complex hybrid systems or behavior based control systems)

Question 3

What does Behavior Arbitration use?

Answer 3

Some decision function (probability, activation spreading, utility)

Question 4

Synonym for Behavior fusion

Answer 4

Cooperative method

Question 5

Behavior Fusion
Cooperative method
1. What is needed to make it possible?
Example
2. Which level of representation is it applied to?

Answer 5

1. The behaviors representations must be combinable
   - > velocity (easy to sum up)
2. Lower-level representations of actions (directions, velocity)

Question 6

How to avoid oscillation between two competing behaviors (2)?

Answer 6

Weigh behaviors, some have stronger influence on the final result + logic to prevent certain combinations/outcomes

Question 7

Behavior Arbitration

1. Which level of representation is it applied to?
2. Can it be combined with behavior fusion?

Answer 7

1. High-level representations (turn, stop, grasp)

2. Yes, on different levels

Question 8

Emergent behavior

1. Definition
   - >
2. When does it get unpredictable (2)?

Answer 8

1. Emergent behavior is structured (meaningful) behavior that is apparent at one level of the system (observer viewpoint) but not at another (from controllers viewpoint)
   - > frame of reference problem (perspective issue)
2. When environment is dynamic or controller has interacting components

Question 9

What is needed to call something emergent behavior (2)?

Answer 9

1. External observer (to observe and describe behavior (&possibly be surprised))
2. Access to the controller (to check if the emerging behavior is not pre-specified)

Question 10

Navigation

1. Path planning problem

Answer 10

1. Map: Yes
2. Target location: Yes
3. Own location: Yes
   - > needs to plan and follow the path to target location

Question 11

Navigation

2. Localization Problem

Answer 11

1. Map: Yes
2. Target location: Yes
3. Own location: No
   - > needs to find location on map

Question 12

Navigation

3. Coverage problem

Answer 12

1. Map: Yes
2. Target location: No
3. Own location: Yes
   - > robot needs to check all parts of the map to find the target (cover whole area)
   - > can be done without map, but is really difficult

Question 13

Navigation

4. Mapping problem

Answer 13

1. Map: No
2. Target location: Yes
3. Own location: Yes
   - > robot needs to build a map as he goes

Question 14

Navigation

5. SLAM problem

Answer 14

1. Map: No
2. Target location: No
3. Own location: No
   - > Simultaneous Localization and Mapping (SLAM)

Question 15

Which two problems are essentially the same?

Answer 15

Path planning and coverage, only with a known, single goal.

Question 16

Localization

1. Simple way to measure?
2. What is using this measure called?
3. When does it become inaccurate?

Answer 16

1. Odometry
2. Path integration
3. The further the robot goes

Question 17

Localization

Solution for odometric error

Answer 17

Use measurements of the environment and compare them with the map data.

Question 18

Localization

1. Name of representation for robots position

Answer 18

1. Belief representation

Question 19

Localization
Belief representation - advantage disadvantage
1. Single-hypothesis belief
2. Multiple-hypothesis belief

Answer 19

1. A: no position ambiguity, D: ambiguous observations

2. A: Multiple possible positions are tracked, D: position ambiguity

Question 20

Localization
Multiple-hypothesis belief
1. Particle Filter
2. What do they simulate (4)

Answer 20

1. Represents multiple possible positions of the robot

2. Position, Movement (change in pos), observations, noise

Question 21

Localization
Kalman Filter
1. What does it do?
2. What is related?
example

Answer 21

1. It combines multiple measurements with different (un)certainties
2. It relates the landmark positions with the robots position certainty
   - > more certainty about landmark = more certainty about robots position

GPS combined with odometry of car