Optimizing Warehouse Robot Navigation with LSTM Networks and Adaptive Greedy Weight Techniques

As warehouse automation continues to advance, effective path planning for robots in dynamic and complex environments is becoming increasingly important. Traditional path planning algorithms have limitations when it comes to handling dynamic obstacles and real-time changes. To enhance robots' navigation capabilities in such environments, this study combines Long Short-Term Memory (LSTM) networks with adaptive greedy weight strategies to propose an optimized path planning method. LSTM networks can predict the future paths of robots based on historical data, while the adaptive greedy weight strategy balances exploration and exploitation during path selection. This approach improves the efficiency and safety of robot navigation in complex warehouse settings.

The dataset contains comprehensive source codes, including the implementation of an LSTM model, LSTM training processes, and two advanced pathfinding algorithms: one that combines LSTM with adaptive greedy weight strategies, and another that integrates the A* algorithm with adaptive greedy weight. These codes are designed to facilitate research in dynamic path planning, allowing for the exploration of sophisticated optimization techniques that enhance robot navigation in complex, real-time environments. The inclusion of LSTM training code provides a foundation for developing and fine-tuning the model, enabling a deeper understanding of how machine learning-based approaches can improve predictive accuracy and adaptability. Additionally, this dataset supports comparative analysis between traditional and machine learning-enhanced algorithms in handling dynamic obstacles and real-time path adjustments, making it a valuable resource for researchers and practitioners in the fields of autonomous systems and robotics.