TRAINING NEURAL-NET CONTROLLERS WITH THE HELP OF TRAJECTORIES GENERATED WITH FUZZY RULES (DEMONSTRATED WITH THE TRUCK BACKUP TASK)

In controls, it is known that control path planning is a task quite di fferent from that of ''next-step'' control action generation. This is, for example, why training a neural-net controller for the truck backu p task is so tedious; the appropriate backing trajectories have to be discovered by trial and error before the net can be depended on to beh ave well in a ''next-step'' manner. In this paper we suggest and demon strate that this awkwardness can be circumvented by supplying a certai n amount of overall system knowledge with use of trajectories (system paths) generated with fuzzy logic, and training the neural-net control ler to learn how to generate next-step control actions more or less in conformance with those trajectories. The fuzzy rules and membership f unctions need not be optimum. The neural-net controller trained in thi s way performs in a manner superior to that of the fuzzy controller. W e demonstrate these circumstances with the truck backup task, using a functional-link net, and making use of only 4 trajectories generated w ith fuzzy rules. (C) 1998 Elsevier Science B.V. All rights reserved.