Resource-Aware Design for Reliable Autonomous Applications with Multiple Periods

Reliability is the most important design issue for current autonomous vehicles. How to guarantee reliability and reduce hardware cost is key for the design of such complex control systems intertwined with scenario-related multi-period timing behaviors. The paper presents a reliability and resource-aware design framework for embedded implementation of such autonomous applications, where each scenario may have its own timing constraints. The constraints are formalized with the consideration of different redundancy based fault-tolerant techniques and software to hardware allocation choices, which capture the static and various causality relations of such systems. Both exact and heuristic-based methods have been implemented to derive the lower bound of hardware usage, in terms of processor, for the given reliability requirement. The case study on a realistic autonomous vehicle controller demonstrates the effectiveness and feasibility of the framework.