Obstacle Avoidance Prediction System of Hospital Distribution Robot Based on Deep Learning

Abstract

With the global outbreak of COVID-19, the normalization of epidemic is a serious fact that people have to deal with. In this paper, obstacle avoidance prediction system based on deep learning is designed to forecast and avoid obstacles. The system firstly is used the bidirectional cyclic neural network to predict pedestrian trajectory to generate dynamic obstacle information. With dynamic obstacle information into global information, the hospital distribution robot with the system can load static obstacle information and merge. The effectiveness of the system is verified through simulation experiments in two scenarios of complex areas and straight corridors in the hospital. Simulation experiments show that it realizes global path planning through dynamic window method. The obstacle avoidance prediction system of this paper can avoid direct contact between doctors and mild patients and reduce the workload of medical staff and the risk of infection effectively.