Adaptive Task Migration for Peer-Level Dual-MCU Embedded Robotics Systems

Peer-level dual-MCU architectures partition real-time control and computational workloads to improve timing determinism in embedded robotic systems (Johnson, 2025). However, static task allocation does not account for dynamic workload variations, temperature spikes, or unpredictable sensor input, which can introduce temporary latency spikes and reduced throughput. This letter presents an adaptive task migration framework that continuously monitors CPU load, thermal conditions, and inter-MCU latency to redistribute tasks in real time without violating control-loop deadlines. Experimental results on two dual-MCU platforms demonstrate a 12% increase in throughput, 23% fewer latency spikes exceeding 5 ms, and an 8.3 °C reduction in peak coordinator MCU temperature. Task migration overhead remains below 580 µs, preserving deterministic control performance.