Deep Reinforcement Learning and Proximal Policy Optimization for Jetbot Automation

In this work explores the complex design of a state-of-the-art deep reinforcement learning autonomous navigation system that is designed to optimize object delivery in industrial warehouse environments by utilizing Proximal Policy Optimization (PPO) techniques. The system is controlled by the car-shaped robot JetBot, which is equipped with a cutting-edge NVIDIA AI-focused board. It is carefully designed to maneuver around warehouse conditions with remarkable efficiency. Among its primary functions are the ability to move past various obstacles with ease and to enable the automatic delivery of objects to warehouse staff. Extensive in-lab experiments carried out in virtual warehouses accurately mirrored real-world situations, integrating a wide range of impediments like boots, gloves, tools, and implements to fully evaluate the system's capabilities. By means of carefully selected trials, the system's ability to navigate complex warehouse layouts and quickly detect and avoid obstacles was thoroughly tested. The outcomes demonstrated the system's resilience and flexibility in negotiating changing warehouse settings, signaling a major advancement in autonomous warehouse logistics.