Dynamics Modeling for Autonomous Container Trucks Considering Unknown Parameters

Autonomous container trucks (ACTs) contribute significantly to transportation efficiency in ports. Their dynamics modeling is an indispensable component for constructing advanced driving protocols. However, the presence of unknown dynamics terms substantially constrains the modeling accuracy, which is attributed to the uncertainty of the external loads and operating conditions of the ACTs. This study proposes a novel adaptive particle swarm optimization algorithm with a levy flight strategy (APSO-Levy) to calibrate the unknown terms and construct a three-degree-of-freedom dynamics model. Adaptive particle swarm optimization (APSO), standard particle swarm optimization (PSO), and four recent PSO variants are considered baselines. We calibrate the dynamics model for two driving scenarios (straight-line driving and mixed driving) across three load conditions (no load, half load, and full load) based on the aforementioned seven algorithms. A total of 42 calibrations are conducted. The experimental data from real ACTs in Xiamen Port, China, are employed for parameter calibration. According to the results, the calibration performance of the APSO-Levy is the best, with average fitness being 27.627% and 11.939% better than the standard PSO and the APSO, respectively. Meanwhile, the proposed algorithm outperforms the four variants. These findings highlight that the APSO-Levy is the recommended approach for calibrating the dynamics model of the ACTs. The effectiveness of the calibration is demonstrated by comparing the actual observations with the acceleration values from the dynamics model.