Dynamic Adaptation of Individual Perception-Action Control Plans in a Heterogeneous Team of Intelligent Mobile Agents
Abstract
The objective of this project was to explore the development of new type of physical agents, called autonomous observers. An autonomous observer is a mobile robot equipped with cameras that can perform vision tasks in response to high level inputs given by human users. To perform such tasks, multiple observers may team to reach the same results quicker or to attain goals that no agent could achieve alone. For example, finding and/or tracking a fast target reliably in a complex environment may not be possible with a single observer. Autonomous observers allow users to perform remote observation tasks without worrying about the details of camera motions. Instead, these motions are automatically computed and executed. Our research has addressed three major topics: model building, target seeking, and target tracking. This sequence of three topics is implicitly based on the following hypothetical scenario: autonomous observers are dropped into an unknown environment, of which they first have to build a model; then they have to find a smart target hiding among view obstructing obstacles; finally, they have to monitor this target and track its motions. For each topic, we have developed and implemented new algorithms. We have experimented with our software both in simulation and with an autonomous observer prototype that we have designed and built.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1997
- Accession Number
- ADA329458
Entities
People
- Jean Claude Latombe
Organizations
- Stanford University