Social Robot Navigation and Comfortable Following: A Novel Knowledge-Based Robot-Pedestrian Interaction Model with Self Learning Strategy

Currently, mobile service robots have been playing an increasingly significant role in daily life. However, many mobile service robots still face challenges in navigation and interaction modes. Current navigation and interaction patterns are often fixed and lack personalization, making it difficult to meet the preferences of different users. Users often need to continually provide specific action instructions to the robot, resulting in a complex and cumbersome interaction process. This paper aims to propose a human-robot interaction framework for robot navigation and pedestrian following based on natural language processing (NLP), which combines a knowledge base and neural network predictions for comfortable following distance to achieve a more intelligent, comfortable, and personalized human-robot interaction experience. To validate the effectiveness of the proposed framework and algorithms, a series of experiments were conducted in different scenarios. The results demonstrate that the NLP-based navigation and following human-robot interaction framework presented in this paper exhibits favorable feasibility and applicability in various working environments. Additionally, the neural network-based algorithm for predicting comfortable following significantly enhances the pedestrian following experience.