Finite budget annealing for predictive trajectory control of four wheel steering cable handling vehicles
Four-wheel-steering automated guided vehicles require accurate trajectory control with explicit actuator and geometric limits. We construct a fully specified numerical benchmark coupling a fourteen-state wheel and chassis plant to a reduced model predictive controller. Standard simulated annealing, chaos-initialized annealing, deterministic predictive control and a documented PID baseline are compared in four scenarios. Forty paired horizon searches use an equal budget of 300 objective evaluations, alongside twelve plant-perturbation cases and twenty-five payload–friction runs. The median final objective is 1.34903 for standard annealing and 1.33574 for the chaotic method, which wins 22 of 40 pairs; the paired uncertainty interval includes zero. Sinusoidal cross-track errors are 0.0222 m for chaos-initialized control and 0.0272 m for PID, but tracking, speed settling, electrical input and computation do not share a uniform ranking. Deterministic predictive control provides the smallest obstacle-route tracking error, while the chaotic method incurs greater modeled electrical input than PID. Archived parameters, code, trajectories and optimizer histories connect each result to an executable implementation. The findings support a conditional accuracy–actuation–computation tradeoff rather than a general advantage from chaotic initialization.
Authors
- Fenqian Wu (ORCID: https://orcid.org/0009-0005-7712-6490)
- Wenliang Liu (ORCID: https://orcid.org/0000-0002-9455-1369)
- Yibin Zheng¹
- Weiyuan Li²
- Zhixiong Lin¹
Institutions
- State Grid Fujian Electric Power Company (China) (CN)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-18
- DOI
- https://doi.org/10.21203/rs.3.rs-11512991/v1
- Primary Topic
- Vehicle Dynamics and Control Systems
- Type
- preprint