Planning with a functional neural-network architecture

This article introduces the concept of planning in an interactive environme nt between two systems: the challenger and the responder, The responder's t ask is to produce behavior that relates to the challenger's behavior throug h some response function. In this setup, we concentrate planning on the res ponder's actions and use the produced plan in order to control the responde r. In general, the responder is assumed to be a nonlinear system whose inpu t-output (I/O) map may be expressed by a Volterra series. The planner uses an estimate of the challenger's future output sequence, the response functi on, and a model of the responder's I/O relation implemented through a funct ional artificial neural network (FANN) architecture, in order to produce th e input sequence that will be applied to the responder in the future, in pa rallel-time with the challenger's corresponding output sequence. The respon der accepts input from the planner, which may be combined with feedback inf ormation, in order to produce an output sequence that relates to the challe nger's output sequence according to the response function. The importance o f planning for the generation of smooth behavior is discussed, and the effe ctiveness of the planner's implementation using neural network technology i s demonstrated with an example.