Autonomous Vehicle Mission Planning Using AI (Artificial Intelligence) Techniques.
Abstract
This thesis investigates a software architecture for autonomous vehicle control. The autonomous vehicle's planning ability is divided into operating system functions and mission planning system functions. The blackboard control architecture is adopted for the operating system design with implementation using the ROSS programming language. The planning system incorporate elements of a plannar land understander by declaratively encoding meta-knowledge, or knolwedge about the planning process. By separating the knowledge about how to plan from the specific domain knowledge, an understander can use this knowledge about how plans are constructed, in combination with the specific domain knowledge, in the understanding process. Likewise, a planner can use this same knowledge in the planning process. Thus, a great deal of flexibility is strained by dividing the knowledge base into meta-rules and omain specific rules. The planning system constructs an agenda of scripts which directs the control flow in the operating system. The operating system is given the additional duties of goal detector and plan projector and detects any errors in the plan. The implementation demonstrates the benefits of using meta-planning concepts combined with a blackboard control architecture to provide an autonomous vehicle with a more flexible and powerful planning capability.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1985
- Accession Number
- ADA163956
Entities
People
- Samuel E. Stockbridge
Organizations
- Air Force Institute of Technology