Visible to the public Thermal Management of Cyber-Physical Systems


Modern cyber-physical systems are monitored and controlled by multi-core platforms, and thermal management of multi-core chips is critical as overheated cores thereon will suffer from exponentially decaying lifetime and unacceptable performance degradation. To meet the timing and system lifetime reliability requirements under dynamic workloads and operating environment, we need a real-time thermal management (RTM) scheme that predicts run-time temperature and actuates effective thermal control without compromising task deadlines. For real-time thermal prediction in vehicle ECUs (ECU RTM), we propose analytic thermal prediction model combining thermal characteristics of multi-core platform and application task workload, and also develop a method for run-time temperature estimation with low overhead using performance counters. For effective thermal control in mobile systems (Mobile RTM), we consider thermal coupling between processors, external cooler, and battery, and then optimize thermal control knobs based on thermal prediction methods. For optimization of multi-objective mission performance in UAV systems, we model thermal dependency to minimize thermal degradation of system componenst. For fault-tolerance in an automated highway system, a resource-aware framework for adaptive fault-tolerance (AdaFT) is proposed for distributed CPS. These joint efforts in thermal management of CPS achieved 32% utilization improvement and 63% lifetime prolongation in vehicle ECUs and 13 % real-time performance improvement in mobile systems over state-of-the-art dynamic thermal management.

Creative Commons 2.5

Other available formats:

Thermal Management of Cyber-Physical Systems
Switch to experimental viewer