Motion Planning for Intelligent Robots Using LSTM-Based Time Series Prediction and Dueling DQN

Over the past few decades, motion planning algorithms have advanced significantly across various scientific and technological fields. Notable improvements in path planning and trajectory optimization techniques have been made possible by sampling-based techniques and convex optimization strategies, among other things. In recent years, there has been a notable surge in the usage of robot manipulators in scientific research and engineering applications. There has been development of technologies like physics-based simulation, autonomous robotics and the application of artificial intelligence methods. This paper proposed an innovative model that integrates machine learning approaches to improve the motion planning of intelligent robots with time series prediction using LSTM. With minimal tracking changes, the dueling DQN approach generates a smooth path. The motion sensor data from the intelligent robot are utilized to predict the future state of the mobile robot. This paper addresses various challenges to motion planning under different environmental conditions with or without collision objects. Combining OpenAI Gym with the physics-based simulator PyBullet demonstrates adaptability and room for growth. Compared to the traditional systems, the proposed model outperforms efficiently in finding the optimal path with proper planning.