LLA-MPC on Embedded Hardware: Rapid Adaptive Control with Thousands of Parallel Models

This research project explores the implementation and evaluation of Look-Back and Look-Ahead Adaptive Model Predictive Control (LLA-MPC) on resource-constrained embedded hardware. Designed as a learning-free framework, LLA-MPC enables rapid, real-time system identification and control by evaluating thousands of parallel candidate models simultaneously. While originally formulated in simulation for autonomous racing, this methodology adapts the framework for physical deployment under noisy state estimation and constrained computational budgets. The implementation features modular dynamics and integrators, facilitating online parameter identification—such as varying surface friction or environmental boundary changes—in real time. For environmental engineering applications, this adaptive control paradigm offers significant potential in optimizing mobile robotic sensing platforms for pollutant tracking, or managing dynamic

actuator responses in smart indoor air quality control systems. The research focus centers on evaluating the computational efficiency, tracking accuracy, and adaptation speed of the parallel-model framework compared to conventional fixed nominal controllers. Through systematic simulation and hardware-in-the-loop testing, the project provides implementation support and research direction for deploying high-throughput adaptive control algorithms on embedded edge devices.

Message: Hi FE, I need help with “LLA-MPC on Embedded Hardware: Rapid Adaptive Control with Thousands of Parallel Models” in “Environmental Engineering”