SSS-DDQN solution-space best-first search for robust trajectory tracking of Mecanum robots under partial...

Robust trajectory tracking of Mecanum-wheel mobile robots under partial observability is challenging due to nonlinear motion coupling, noisy measurements, sensor dropout, and external disturbances. In this study, we propose an SSS*-DDQN solution-space best-first search controller for robust trajectory tracking of a four-Mecanum-wheel mobile robot. The proposed framework integrates extended Kalman filter (EKF)-based belief-state estimation, QMDP belief sampling, DDQN value evaluation, and SSS* OPEN-list search with LIVE/SOLVED node refinement. The EKF estimates the hidden robot state under noisy observations and dropout, the DDQN provides learned long-term value evaluation, and the SSS* search refines promising corrective-action clusters using beta-bound pruning, delayed MIN-sibling refinement, bounded OPEN-list management, and secondary lookahead. The controller is evaluated using lemniscate trajectory tracking and compared with EKF-belief PER-DDQN, QMDP-PER-DDQN, Alpha-Beta Minimax, and AB-DDQN Hybrid controllers. The proposed SSS*-DDQN achieved the best overall ranking with ${\text{RMSE}}_{xy}=0.01906$ m, ${\text{MAE}}_{xy}=0.016074$ m, and a cumulative reward of 1843.6. Compared with the AB-DDQN Hybrid baseline, it reduced ${\text{RMSE}}_{xy}$ by 12.93% and ${\text{MAE}}_{xy}$ by 17.80%. Statistical validation over multiple stochastic trials further showed the lowest mean ${\text{RMSE}}_{xy}$ of $0.01847\pm 0.00132$ m and ${\text{MAE}}_{xy}$ of $0.01554\pm 0.00128$ m, with narrow 95% confidence intervals. The results confirm that the proposed SSS*-DDQN improves robust action selection and provides consistent tracking performance under partial observability.