Approximate LASSO Model Predictive Control for Resource-Constrained Systems
Published in 2020 Sensor Signal Processing for Defence Conference (SSPD): Proceedings, 2020
This paper addresses the challenges of deploying LASSO-based Model Predictive Control (MPC) on resource-constrained systems, such as embedded platforms, due to the intensive memory usage and computational cost as the horizon length increases. The authors propose a reduced-precision implementation of the Lean Proximal Gradient Descent (LPGD) algorithm, tailored for FPGA platforms.
Key contributions include:
Reduced Precision Approximation: The LPGD algorithm is adapted to operate with reduced precision, significantly lowering hardware resource requirements.
FPGA Implementation: The approximate LPGD solver is implemented on an FPGA, demonstrating substantial reductions in logic cost (up to 60%), memory bandwidth (up to 80%), and power consumption (up to 70%), while maintaining comparable control performance to high-precision solvers.
Reconfigurable Design: The proposed design is reconfigurable, allowing for adaptability to different system requirements and constraints.
Recommended citation: Wu, Y., Mota, J. F. C., & Wallace, A. M. (2020). Approximate LASSO Model Predictive Control for Resource-Constrained Systems. In *Proceedings of the 2020 Sensor Signal Processing for Defence Conference (SSPD)*. IEEE. https://doi.org/10.1109/SSPD47486.2020.9272000
Download Paper | Download Slides | Download Bibtex