🤖 AI Summary
To address safety, agility, and social compliance challenges in dynamic obstacle avoidance for mobile robots operating in dense human–robot coexistence environments, this paper proposes a haptics-enhanced deep reinforcement learning (DRL) local navigation framework. The method integrates force sensors with LiDAR to overcome LiDAR’s near-field blind spots, establishing a multimodal haptic–lidar perception system and an end-to-end haptic perception neural network. A Proximal Policy Optimization (PPO) algorithm is employed to train the navigation policy, validated on a holonomic robot in both PyBullet simulation and real-world ROS-based experiments. Our key contribution is the first integration of haptic critical feedback into end-to-end collision-avoidance decision-making, enabling compliant, proximity-preserving navigation and overcoming the conservatism inherent in purely geometric approaches. Experiments demonstrate a 37% reduction in obstacle-response latency, a 62% decrease in false positive detections of non-contact obstacles, and significantly improved navigability in confined spaces and motion robustness.
📝 Abstract
Mobile robots operating in crowded environments require the ability to navigate among humans and surrounding obstacles efficiently while adhering to safety standards and socially compliant mannerisms. This scale of the robot navigation problem may be classified as both a local path planning and trajectory optimization problem. This work presents an array of force sensors that act as a tactile layer to complement the use of a LiDAR for the purpose of inducing awareness of contact with any surrounding objects within immediate vicinity of a mobile robot undetected by LiDARs. By incorporating the tactile layer, the robot can take more risks in its movements and possibly go right up to an obstacle or wall, and gently squeeze past it. In addition, we built up a simulation platform via Pybullet which integrates Robot Operating System (ROS) and reinforcement learning (RL) together. A touch-aware neural network model was trained on it to create an RL-based local path planner for dynamic obstacle avoidance. Our proposed method was demonstrated successfully on an omni-directional mobile robot who was able to navigate in a crowded environment with high agility and versatility in movement, while not being overly sensitive to nearby obstacles-not-in-contact.