Thermal-Aware Approximate Real-Time Task Scheduling for Multicore Embedded Systems

High temperature affects the processing performance of chips and accelerates electronic component aging, leading to reliability issues. This paper proposes a two-stage temperature-aware scheduling scheme for approximate tasks based on homoge-neous multicore platforms, solving approximate task scheduling problem under time, energy, reliability, and temperature constraints. In the first stage, we design a heuristic scheduling algorithm to perform task-to-processor assignment and preliminarily task frequency selection to minimize energy and makespan. In the second stage, we design a temperature optimization algorithm that combines DVFS and slack distribution to ensure that processor temperature is lower than the temperature threshold. Compared with the existing methods, the experimental simulations show that our method achieves 97.6% runtime reduction and 14.9% peak temperature reduction.