Optimization Methods for Computing System in Mobile CPS

Abstract

In Mobile Cyber-physical system (CPS) is a popular research field in recent years. It aims to control and monitor mobile devices in complex and real-time scenes, and provide people with convenience and economy by using intelligent applications. The scenes in mobile CPS have close relationships with everyone's life and it has pervasive effect on people's life anytime and anywhere. In the paper, we focus on the intelligent transportation systems (ITS) in mobile CPS. Specifically, we analyse the bottlenecks in automated and autonomous vehicles and find the on-car computer is the key component for low latency and low power self-driving vehicles. We show detailed research developments on the most advanced CPU-GPU platform including four aspects, mainly putting focus on the methods to reduce latency and energy in the system. The four aspects cover reducing data transferring overhead, reducing power consumption with power-gating, optimizing warp scheduling scheme and optimizing cache performance in coprocessors. Finally, we give the prospects and development trends in the computing platform of self-driving on-car system.