Hierarchical predictive control for ground-vehicle maneuvering

Berntorp, Karl; Magnusson, Fredrik · 2015 · Crossref

DOI: 10.1109/acc.2015.7171154

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of improving trajectory generation and control for autonomous ground vehicles, particularly during aggressive maneuvers. The authors propose a hierarchical predictive control structure that integrates nonlinear vehicle dynamics at the high level to ensure physical coupling with the low-level actuator allocation. Existing approaches often rely on simplified point-mass models or linearized dynamics at the high level, which can generate reference trajectories that are infeasible for the low-level controller to track, especially when combined-slip tire behavior is significant. The proposed method aims to bridge this gap by using a nonlinear single-track model for high-level planning and a combined nonlinear and linearized model-predictive control (NMPC/LMPC) scheme for low-level execution. The methodology employs a two-layer control architecture. The high-level optimizer uses a nonlinear single-track model with a weighting-function tire model to generate reference trajectories for position, velocity, yaw angle, and yaw rate. This optimization minimizes lane deviation and vehicle sideslip angle while respecting road constraints. These references are passed to a low-level NMPC, which uses a more complex double-track model to allocate individual wheel torques and steer angles. To address the computational burden and potential convergence failures of the nonlinear optimization, a linearized MPC (LMPC) serves as a backup controller. If the NMPC fails to converge within a specified time, the LMPC takes over, ensuring real-time feasibility. The system is implemented using JModelica.org and solved via IPOPT, with symbolic transformations used to reduce computational complexity. Simulation results on a road segment with a 30-meter curvature radius demonstrate the effectiveness of the proposed approach. The combined NMPC/LMPC structure achieved significantly better reference tracking compared to a standalone LMPC approach. The LMPC-only controller tended to overestimate available lateral tire forces, resulting in less aggressive steering and poorer tracking performance. In contrast, the hierarchical approach maintained tighter adherence to the reference trajectory. Although the NMPC experienced three convergence failures out of 77 optimizations in the test scenario, the LMPC backup ensured continuous control. The study confirms that incorporating nonlinear dynamics at the high level improves the physical coupling between planning and control, enhancing performance in aggressive maneuvering scenarios while maintaining feasibility for online implementation.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• steering pattern

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model