Parallel Branch Model Predictive Control on GPUs

2025-06-16

AI Summary

To address the low computational efficiency of Branch-Model Predictive Control (Branch-MPC) under large prediction horizons and multi-scenario settings, this paper introduces the first GPU-accelerated parallel solver tailored for Branch-MPC. Methodologically, it pioneers fine-grained parallelism across both the temporal dimension (prediction steps) and the scenario dimension (branch tree), leveraging a tree-structured sparsity pattern to design a parallel scan algorithm, developing an iterative Linear Quadratic Regulator (iLQR) framework adapted to tree topologies, and integrating the augmented Lagrangian method to handle general inequality constraints. Experimental results on representative autonomous driving tasks demonstrate order-of-magnitude speedups over CPU-based implementations; notably, the solver remains competitive even on short-horizon, small-scale problems. This work significantly advances the real-time solvability of Branch-MPC, enabling its practical deployment in dynamic, uncertainty-rich environments.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Mixed Discrete/Continuous Planning

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

We present a parallel GPU-accelerated solver for branch Model Predictive Control problems. Based on iterative LQR methods, our solver exploits the tree-sparse structure and implements temporal parallelism using the parallel scan algorithm. Consequently, the proposed solver enables parallelism across both the prediction horizon and the scenarios. In addition, we utilize an augmented Lagrangian method to handle general inequality constraints. We compare our solver with state-of-the-art numerical solvers in two automated driving applications. The numerical results demonstrate that, compared to CPU-based solvers, our solver achieves competitive performance for problems with short horizons and small-scale trees, while outperforming other solvers on large-scale problems.

Problem

• Develop GPU-accelerated solver for branch MPC
• Enable parallelism across prediction horizon and scenarios
• Handle inequality constraints via augmented Lagrangian method

Innovation

• GPU-accelerated parallel solver for MPC
• Temporal parallelism via parallel scan
• Augmented Lagrangian for inequality constraints