A neural network based intelligent planner for the coordinated hybrid agent framework

Abstract

Recently, a coordinated hybrid agent (CHA) framework was proposed for the control of multiagent systems (MASs). In the past few years, it has been applied to both homogeneous and heterogeneous multi-agent systems. In previous studies, the coordination among agents were implemented based on the designer's knowledge of the system. For large complex systems, it would be desirable if we can plan the coordination among agents dynamically. In this study, we demonstrate that an intelligent planner can be designed for the CHA framework to automatically generate desired actions for multiple robots in a multiagent system. The proposed intelligent planner is based on the construction of a biologically inspired neural network. The dynamics of each neuron in the topologically organized neural network is characterized by a shunting neural equation. A landscape of the neural activities for all neurons of a CHA agent contains information about the agent's local goal, permanent obstacles and temporary obstacles. Each agent treats other robots as moving obstacles. The objective for building the intelligent planner is to plan actions for multiple mobile robots to coordinate with others and to achieve the global goal while each agent achieves its local goal. Simulation results show that in order to control a large complex system, an intelligent planner can be designed for the CHA framework so that coordination among agents can be achieved.