Dynamic thermal management for networked embedded systems under harsh ambient temperature variation

Modern vehicle electronics control units (ECUs) are getting rapidly complicated because of active safety and semi-autonomous driving controls, such as electric stability program (ESP) and adaptive cruise control (ACC). Furthermore, the operational environment of ECUs is extremely harsh, especially in terms of an ambient temperature well exceeding 100°C, which causes a very small temperature headroom. Thus, ECUs require a careful temperature management and high performance at the same time. This paper introduces a dynamic thermal management (DTM) mechanism for networked embedded systems for vehicle ECUs. We exploit a fact that the ambient temperature of each ECU is different and also variable by the temperature of the associated unit while previous DTM considers a fixed ambient temperature as a given constant. We propose a new DTM that controls the rates of tasks with computation migration through the vehicle-area-network to maximize the minimum rate of tasks, which in turn enhances the quality of control and increases the maximum safe vehicle speed. We found that the ambient temperature of each ECU has different behavior for a particular operating condition, and suggest to use a proactive computation migration. We demonstrate dramatic improvement in die temperature control accuracy and average of 12.25% improvements in overall quality of control over distributed reactive DTM for our example cases.