A Hybrid Deep Learning Approach for Dynamic Obstacle Avoidance Mobile Robot.
This work presented a novel approach to dynamic collision avoidance in mobile robots by integrating a hybrid
Deep Deterministic Policy Gradient (DDPG) and Adaptive Neuro-fuzzy Inference Systems (ANFIS) algorithm.
This combined approach aimed to enhance the robot's navigation capabilities in dynamic environments by
leveraging the complementary strengths of both DDPG and ANFIS. The model achieved significant achievements,
including a high efficiency score of 0.97012, a robustness rating of 1 (indicating no collisions during testing),
consistent maintenance of a 0.2-meter safety distance, and a success rate of 97.8%. Additionally, the average
completion time of 5.154 seconds demonstrated its real-time decision-making capability, making it suitable for
time-sensitive applications. The proposed hybrid algorithm improved the robot's obstacle detection and decision making abilities, leading to superior performance in dynamic obstacle avoidance scenarios