A neural network-based robot safety system

This paper presents a new approach for real-time robot safety system based on artificial neural networks. This approach includes a neural network detection unit and a neural network decision unit, implemented at an intermediate and high level of sensory processing, respectively. Both the detection and decision units have been implemented and tested by simulation, both separately and as an integrated unit. The response time of the integrated system measured on the 90 MHz, P5 microprocessor is less than 11 ms, and the correctness of safety decisions is 97%