A methodology to design and prototype optimized embedded robotic systems

We have designed a new electric robotic vehicle controlled by a distributed embedded computer system based on the CAN bus, providing features such as, secured manual driving or autonomous driving, and route planning. These functionalities involve control, image and signal processing algorithms executed under real-time constraints. In order to provide a safe software design and to reduce the development cycle time of such complex systems, we use a methodology called A (Algorithm Architecture Adequation). It is based on graphs and partial order, to modelize the algorithm, the distributed architecture (microprocessors, specific integrated circuits, connected by a network) as well as the implementation of the algorithm on the architecture. The Adequation, based on heuristics, consists in finding the best implementation which satisfies the real-time constraints and minimizes the hardware resources. This allows to automatically generate dedicated executives with dead-lock free communication support.