Distributed reinforcement learning social navigation method based on image hidden variable probability model

Abstract

The invention discloses a distributed reinforcement learning social navigation method based on an image hidden variable probability model. According to the method, the image hidden variable probability prediction model is introduced to replace a traditional decisive prediction model, so that on one hand, the reasonability of the model is enhanced, prediction is closer to a pedestrian motion model with randomness, and on the other hand, the strategy performance is further enhanced by enhancing the exploration capability of the model, and overfitting is avoided. Meanwhile, the method achieves discrimination of a dynamic obstacle by decoupling movement of the mobile robot through an image sequence, a high-layer pedestrian detection module with instability is omitted, and sound migration is realized. In addition, a strategy sharing multi-agent simulation environment is designed to simulate a pedestrian dynamic environment, and the similarity between simulation and real crowd interaction is enhanced. In the environment, multiple agents synchronously perform data acquisition, so that the training time can be further shortened.

Description

technical field [0001] The invention belongs to the field of mobile robot navigation, and in particular relates to a distributed reinforcement learning social navigation method based on an image hidden variable probability model. Background technique [0002] With the rapid development of computer and automation technology, robot technology has gradually penetrated into all walks of life and into people's daily life. They not only assist or replace humans in complex and heavy work in traditional manufacturing, but also gradually replace humans in fields such as entertainment, medical care, and security. Under the trend of surging labor costs, the promotion of robot technology will undoubtedly greatly relieve the accompanying pressure. Therefore, robot technology is gradually becoming a current hot research direction, and service robot, as an important branch, has attracted countless attentions. [0003] As an important member of the robotics field, service robots are posit...

AI Technical Summary

Problems solved by technology

However, the training of this type of algorithm often requires a large amount of interactive data in the simulation environment, which is difficult to meet in the real environment or within a limited time.

Images