Socially Aware Robot Navigation with Collision Prediction and Social Norm Integration

• 1、 School of Automation, Guangdong University of Technology, Guangzhou, Guangdong, 510006

Abstract：To tackle the issues of delayed response and insufficient consideration of pedestrian comfort in existing social force models for dynamic pedestrian avoidance, this paper proposes an improved model integrating collision prediction and social norm, named the Collision-aware Orthogonal Steering Social Force Model (COSFM). First, by introducing predicted collision time, the robot’s ability to anticipate dynamic pedestrian trajectories is enhanced. Second, based on four typical social scenarios—head-on encounters, cross-path traversal, active overtaking, and passive yielding—an orthogonal steering interaction force grounded in pedestrian collision prediction is designed, further optimizing the robot’s avoidance strategy. Simulation results demonstrate that, compared to existing social force models, the proposed COSFM outperforms in navigation success rate, path efficiency, and minimum safe distance. Through a lateral avoidance strategy, this approach not only effectively prevents collisions and improves navigation efficiency but also maintains a reasonable distance from pedestrians, exhibiting excellent social adaptability.

Keywords： Social navigation, social force model, dynamic obstacle avoidance, social norm