Using NN-DEVS Approach for Modelling and Simulation of Imperfect Systems: Application to the Reactive Navigation of Autonomous Robot

Abstract

In this paper, we are interested in the modelling and simulation of imperfect systems in the DEVS context. We want to hybrid the DEVS formalism with artificial neural networks and to propose a new modelling and simulation approach which makes it possible to represent the behavior of imperfect systems. The problematic of our work is the integration into the DEVS formalism of tools from artificial intelligence allowing the representation, manipulation, and processing of imperfect data (imprecise, uncertain). NN-DEVS is a new hybrid approach which allows to extend the classic DEVS formalism. This new approach is effective in uncertain systems where the behavior of the system is stochastic. To validate the proposed NN-DEVS approach, we apply this approach to a complex reactive navigation system of a mobile robot.