Predicting human trajectory with virtual HRI environment

Although the topic of HRI (human-robot interaction) has prevailed for a long time, we still require a great endeavor to achieve the vision which robot collaborates with human friendly. One of the reasons for this is the lacking of security. Since robot cannot have the same inteligence and physical experience as human, at least for now, they won’t communicate with us fluently, and may do harm to human physically. In order to stay clear from these accidents, we need to experiment with a safer environment and make robot more smart which can be done with predicting human trajectory. In this paper a virtual environment which built up for HRI is introduced firstly. In this model, interactive object models and human body model which can be controlled by skeleton tracking algorithm. The code of human body model is published on the Github, hoping to help researchers who do HRI debugging in virtual environment Then we propose human trajectory prediction model using neural network and implement it in real world. The results show our prediction model outperforms other typical models and the video of our experiment provides more intuitive perspective.