Cooperative Navigation of Differential-Drive Mobile Robots in Crowd Environments

Mobile robotics has gained a lot of attention due to its wide range of applications in various domains such as warehouses, factories, and hospitals. In crowded environments, cooperative navigation is essential to ensure the safe and efficient movement of mobile robots. The cooperative navigation strategies must enable robots to coordinate their movements and adapt to the dynamic environment. In recent years, significant progress has been made in developing cooperative navigation algorithms for mobile robots. The algorithms can be grouped into centralised and decentralised methods, each with its advantages and disadvantages. However, most deep-learning-based methods have been designed for holonomic robots, while differential-drive mobile robots pose more challenges due to their limited degrees of freedom. The paper presents a method that uses deep reinforcement learning to enable differential-drive mobile robots to navigate cooperatively in crowded environments. The method employs a cooperation module to model the interactions among the robot and other agents, and a self-attention mechanism to prioritize the interaction features. The experimental outcomes show that the method outperforms several existing algorithms, which obtains better success rate and lower collision rate, indicating its effectiveness in guiding robots to their destinations while avoiding collisions.