Spinoza demonstrates a sophisticated real-time vision system for control of a responsive mobile robot, operating in dynamic environments. It is a complex system, coordinating multiple distributed, concurrent processes on a wide range of architectures, and performing a range of asynchronous activities. Its design represents the resolution of conflicting design requirements: high processing capability and telerobotic guidance, under the limitations of a mobile robot: power, heat, space, and communication bandwidth.
The performance gap between specialized digital signal processors and standard personal computers is rapidly shrinking. Therefore the tradeoffs of using embedded hardware versus conventional computer hardware need to be closely examined. Our experience suggests that the costs, in terms of development time and debugging time, are restrictively high using C40s and Transputers. In the future we plan to experiment with Intel based CPUs for vision processing as well as robot control.