an experimental and simulation analysis of multi-objective techniques for mobile robots using improved deep q-network algorithm

Mobile robots have garnered significant attention across various domains, including industrial automation, healthcare, logistics, and autonomous vehicles. however, effective navigation in dynamic and complex environments remains a critical challenge. this research introduces an improved deep q-network algorithm for learning-based mobile robot navigation, addressing a multi-objective optimization problem that seeks to minimize path distance, energy consumption, and travel time within a grid-mapped complex environment. the deep q-network algorithm was enhanced to improve its efficiency in determining the optimal path to a target point. experimental validation using a learning robot demonstrated the effectiveness of the proposed approach, ensuring safe path generation with collision avoidance, optimized path distance, and practical implementability in mobile robot applications. furthermore, training, simulation, and analysis results revealed less than 2% deviations between simulation and experimental outcomes, with a path distance error of only 1.3765%. finally, the proposed algorithm was benchmarked against existing approaches, including the a* algorithm, enhanced deep q-network, and dueling double deep q-network, showcasing its superior performance.