MATLAB source code supporting the research article “QMDP-PER-DDQN Trajectory Tracking for Mecanum Robots under Partial Observability.” This repository provides the MATLAB implementation of a belief-aware trajectory-tracking framework for a four-Mecanum-wheel mobile robot operating under noisy and partially observable sensing conditions. The proposed controller integrates Extended Kalman Filter (EKF)-based Gaussian belief estimation, QMDP belief-sampled action evaluation, Prioritized Experience Replay (PER), and Double Deep Q-Network (DDQN) learning. The repository includes implementations for the proposed QMDP-PER-DDQN controller and the principal baseline controllers, including Full-state PER-DDQN, Noisy-observation PER-DDQN, EKF-belief PER-DDQN, and Alpha-Beta Minimax. It also contains code for component-wise ablation of EKF, QMDP, PER, and DDQN, multi-run statistical validation, performance metric calculation, and deployment-oriented computational analysis. The simulation considers a four-Mecanum-wheel mobile robot tracking a lemniscate reference trajectory under Gaussian observation noise and temporary measurement degradation. The code includes the robot model, observation model, EKF estimation, belief sampling, reinforcement-learning training, reward calculation, baseline comparisons, statistical analysis, and figure generation required to reproduce the principal computational results of the study. MATLAB R2021b or later is recommended. The Deep Learning Toolbox is required. Instructions for running the code are provided in the included README file.
Oorappan G M, Lakshmanan S A, S Vignesh· Zenodo (CERN European Organi...· 0 citations
MATLAB source code supporting the research article “QMDP-PER-DDQN Trajectory Tracking for Mecanum Robots under Partial Observability.” This repository provides the MATLAB implementation of a belief-aware trajectory-tracking framework for a four-Mecanum-wheel mobile robot operating under noisy and partially observable sensing conditions. The proposed controller integrates Extended Kalman Filter (EKF)-based Gaussian belief estimation, QMDP belief-sampled action evaluation, Prioritized Experience Replay (PER), and Double Deep Q-Network (DDQN) learning. The repository includes implementations for the proposed QMDP-PER-DDQN controller and the principal baseline controllers, including Full-state PER-DDQN, Noisy-observation PER-DDQN, EKF-belief PER-DDQN, and Alpha-Beta Minimax. It also contains code for component-wise ablation of EKF, QMDP, PER, and DDQN, multi-run statistical validation, performance metric calculation, and deployment-oriented computational analysis. The simulation considers a four-Mecanum-wheel mobile robot tracking a lemniscate reference trajectory under Gaussian observation noise and temporary measurement degradation. The code includes the robot model, observation model, EKF estimation, belief sampling, reinforcement-learning training, reward calculation, baseline comparisons, statistical analysis, and figure generation required to reproduce the principal computational results of the study. MATLAB R2021b or later is recommended. The Deep Learning Toolbox is required. Instructions for running the code are provided in the included README file.
Oorappan G M, Lakshmanan S A, S Vignesh· Zenodo (CERN European Organi...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
