1994 Fiscal Year Final Research Report Summary
A Study on Acquisition of Control Strategy and Tactics Using Fuzzy Neural Networks
Project/Area Number |
05805036
|
Research Category |
Grant-in-Aid for General Scientific Research (C)
|
Allocation Type | Single-year Grants |
Research Field |
計測・制御工学
|
Research Institution | Nagoya University |
Principal Investigator |
FURUHASHI Takeshi Nagoya University Faculty of Eng., Dept.of Information Electronics, Associate Prof., 工学部, 助教授 (60209187)
|
Co-Investigator(Kenkyū-buntansha) |
GOHARA Kazutoshi Hokkaido University, Faculty of Eng., Dept.of Applied Phisics, Associate Prof., 工学部, 助教授 (40153746)
UCHIKAWA Yoshiki Faculty of Eng., Dept.of Information Electronics, Prof., 工学部, 教授 (20023260)
|
Project Period (FY) |
1993 – 1994
|
Keywords | Knowledge Acquisition / Collision Avoidance / Fuzzy Control / Fuzzy Neural Network |
Research Abstract |
This research was done to establish the technology for automatic acquisition of human experts' control knowledge using fuzzy neural networks. The investigators acquired strategic and tactical control rules which manipulated steering angle and speed of a moving object for avoiding multiple moving obstacles. In 1993, we showed how to acquire the strategic and tactical control rules directly form the operators' manipulating data using the fuzzy neural networks. The automatically obtained fuzzy rules coincided well with those empirically obtained avoiding rules which steered the moving object to the rear of the moving obstacles. In 1994, for multi-input systems, we proposed the fuzzy inference network(FIN)which suppressed the explosion of number of fuzzy rules, made the computation time for the fuzzy modeling short, made the fuzzy modeling with insufficient data possible, made the obtained rules comprehensible. The FIN was applied to the collision avoidance problem with multiple moving obstacles. Simulation results and experimental results showed the feasibility of the proposed network.
|