A Hierarchical Architecture for BehaviorBased Robots

1
A Hierarchical Architecture for Behavior-Based
Robots

• Monica N. Nicolescu and Maja J. Mataric
• International Conference on Autonomous Agents
  Proceedings of the first international joint
  conference on Autonomous agents and multiagent
  systems

Presented by Yu-Han LuJune 11 2004s9008_at_cherry.c
s.nccu.edu.tw
2
Outline

• Introduction
• Adapting Behavior for Representation
• Behavior Representation
• Hierarchical Behavior Networks
• Experimental Results
• Conclusion

3
Introduction

• Behavior-based system
• Collection of behaviors
• Sensors
• Effectors
• Limitation
• Lack abstract representation
• Reusability

4
Adapting Behavior for Representation

• Behaviors-based Systems Behavior
• Rules
• Taking input from sensors or other Behaviors
• Sending outputs to effectors or behaviors
• Rule
• World preconditions
• Sequential preconditions

5
Abstract Behavior

• Manipulate and change conditions at an abstract
  representation level.
• Behavior are treat as high-level operators.
• Separation of the activation conditions from the
  outputs.
• One abstract behavior can trigger many primitive
  behavior.

6
Adapting Behavior for Representation
7
Behavior Representation

• Abstract Behavior Input
• UseBehavior if used in the current network
• ActivLevel sums the activation messages
• Inhibit whether this is inhibited by another
• Sensory Input a set of inputs
• Preconditions
• Continue not finish, continues sends message
• Abstract Behavior Output
• Active active primitive behavior
• Effects signals the status of postconditions

8
Behavior Representation
9
Behavior Network

• The behavior networks are a means of specifying
  task plans as networks whose nodes are abstract
  behaviors and where the links between them
  represent precondition and postcondition
  dependencies.

10
Behavior Networks

• Components and Structure
• Allow to perform complex temporal sequence.
• Allow behaviors to be reused
• Preconditions
• Permanent preconditions must met during
  execution
• Enabling preconditions must met before the
  execution
• Ordering constraints must have been met before
  the execution

11
(No Transcript)
12
Behavior Network Execution

• Sequential execution
• Enforced by the existence of precondition and
  postcondition.
• order matter
• Opportunistic execution
• Not place temporal dependencies
• order not matter
• Activation level number of successor behaviors
  in the network that require the achievement of
  its postconditions.

13
Hierarchical Behavior Networks

• Flat representations
• Expressive, Flexible
• Not have modularity
• Hierarchical behavior network
• Network Abstract Behavior (NAB)
• Using existing modules
• Postcondition will true when subnet is finished
• Connect Active output to NAB or AB

14
(No Transcript)
15
Experimental Results
16
(No Transcript)
17
Conclusion

• Increasing levels of abstraction
• Reduce complexity

18
Q AThank you!

• Presented by Yu-Han Lu
• June 11 2004
• s9008_at_cherry.cs.nccu.edu.tw