Thermal Coupling as a Source of Sustained Control Degradation in Embedded Robotic Controllers

Sustained execution of mixed real-time control and computational workloads on single microcontrollers often leads to long-term degradation in timing stability. While scheduling interference is commonly cited as a cause, the contribution of thermal effects is less well characterized. This letter experimentally isolates thermal coupling from scheduling effects in an embedded robotic controller executing fixed-priority workloads. Using an ESP32-S3-based platform, control-loop timing is compared under sustained load in a single-MCU configuration and a thermally distributed dual-MCU configuration with identical firmware. Results demonstrate that control-loop drift and interrupt latency degradation in the single-MCU system correlate with thermal throttling rather than scheduler interference. Offloading computational load to a physically separate microcontroller reduces peak junction temperature by up to 28 °C and preserves sustained control-loop stability without modifying control algorithms. These findings indicate that thermal coupling is a significant factor in long-term real-time degradation of embedded robotic controllers.