A reinforcement connectionist path-finder
The problem of path-finding is investigated using a reinforcement connectionist system able to find and learn feasible paths. The learning phase is a stochastic trial and error procedure from performance feedback. These concepts are described and applied with the objective of real-time decision. The paper presents the problem formulation, the system architecture and the learning algorithm. Experimental results are given, including robustness aspects and generalisation capabilities.
Bibliographic Reference: Paper presented: 6th International Conference on CAD/CAM, Robotics and Factories of the Future, London (GB), Aug. 19-22, 1991
Availability: Available from (1) as Paper EN 36334 ORA
Record Number: 199111102 / Last updated on: 1994-12-02
Original language: en
Available languages: en