Neural network middleware for model predictive path tracking of networked mobile robot over IP network

The potential use of IP networks for real-time high performance robots and automation is enormous and appealing. A widely attractive objective for an IP-based mobile robot is to control a mobile robot over the IP network to track a predefined path. This paper proposes a model predictive path tracking control methodology over an IP network via middleware. In addition to the normal use of middleware, this paper utilizes middleware to schedule a control parameter for IP network delay compensation. The parameter is adjusted externally at the output of the path tracking algorithm with respect to the current network traffic conditions and a predictive performance measure computed by a neural network. Simulation results show that the mobile robot with neural network middleware provides significantly better IP networked control system performance.