The National Science Foundation (NSF) awarded a CAREER grant titled "CHRONOSDRIVE: Ensuring Timing Correctness in DNN-Driven Autonomous Vehicles with Accelerator-Enhanced Real-Time SoC Integration" to Wayne State University under the Computer and Information Science and Engineering (CISE) program (CFDA 47.070). The $229,303 project is aimed at developing an integrated architecture that leverages hardware-software co-design to address complex timing and safety challenges for autonomous driving systems and other autonomous machines powered by deep neural networks (DNNs). The key objectives are: 1) developing an accelerator-enhanced system-on-a-chip (SoC) that enables real-time scheduling decisions at the hardware level, 2) designing an accelerator-aware real-time scheduling framework to accurately characterize DNN task timing requirements, and 3) implementing an outlier management strategy to derive reliable worst-case execution time values for DNN tasks on accelerators. This project seeks to establish the foundation for certifiable, safety-critical autonomous machines that deliver assured real-time performance and reliability.
Generated 3/18/25, 4:33 AM