EASYR: <u>E</u> nergy-Efficient <u>A</u> daptive <u>Sy</u> stem <u>R</u> econfiguration for Dynamic Deadlines in Autonomous Driving on Multicore Processors

The increasing computing demands of autonomous driving applications have driven the adoption multicore processors in real-time systems, which turn renders energy optimizations critical for reducing battery capacity and vehicle weight. A typical optimization method targeting traditional systems finds a speed under static deadline, resulting conservative savings that are unable to exploit dynamic changes system environment. We capture emerging deadlines arising from vehicle’s change velocity context an additional opportunity. In this article, we extend preliminary work uniprocessors [ 66 ] processors, introduces several challenges. use state-of-the-art gang scheduling 5 mitigate some However, it entails NP-hard combinatorial problem tasks need be grouped into gangs tasks, formation, could significantly affect saving result. As such, present EASYR, adaptive reconfiguration approach generates given directed acyclic graph dynamically adapts parameters processor speeds satisfy while consuming as little possible. timing constraints also satisfied between reconfigurations through our proposed safe mode protocol. Our extensive experiments with randomly generated task graphs show formation heuristic performs 32% better than one. Using set Bosch real-world data, EASYR achieves reductions up 30.3% on average scenarios compared conventional current demonstrating great potential gains by exploiting deadlines.