Session

DescriptionThis session presents techniques for collaborative, efficient, and adaptive design of CPS and IoT systems. Several papers in the session address the key challenge of collaborative edge-cloud execution of machine learning workloads. Innovative techniques such as multi-objective neural architecture search, complexity-aware selection of inference-location, and bandwidth-saving approaches such as adaptive frame clustering and operating on compressed video streams are proposed. The session also features papers that propose design frameworks to automatically learn neural network controllers and a novel priority assignment approach for CAN bus messages.