Project Grant 2343601

The National Science Foundation (NSF) awarded a $600,000 Project Grant under the Computer and Information Science and Engineering (CFDA 47.070) federal grant program to the University of Delaware. The grant will fund a 3-year research project to develop a framework for predictable deep neural network (DNN) inference in autonomous vehicle (AV) perception systems. The key objectives are to: (1) understand the challenges of timing predictability in DNN inference for AVs, (2) design a framework for predictable DNN inference in multi-sensor, multi-task AV perception, and (3) integrate this framework into the Autoware AV software pipeline. The project aims to mitigate DNN inference time variations and ensure timing predictability for safety-critical AV applications, leveraging techniques such as feature map caching, sensor data fusion, and multi-tenant DNN inference co-scheduling. Comprehensive evaluations will be conducted using open AV datasets, an indoor connected and autonomous vehicle testbed, and a university-based Level-4 autonomous Lincoln MKZ vehicle.