Learning intentions for improved human motion prediction

• Learning human’s goal position from the data using growing hidden Markov models.
• Human motion prediction using social forces-based motion models exploiting estimated goal.
• Extended experimental analysis of both strengths and weaknesses using the existing data set.
• Comparison with benchmark strategy shows significant performance gain.

For many tasks robots need to operate in human populated environments. Human motion prediction is gaining importance since this helps minimizing the hinder robots cause during the execution of these tasks. The concept of social forces defines virtual repelling and attracting forces from and to obstacles and points of interest. These social forces can be used to model typical human movements given an environment and a person’s intention. This work shows how such models can exploit typical motion patterns summarized by growing hidden Markov models (GHMMs) that can be learned from data online and without human intervention. An extensive series of experiments shows that exploiting a person’s intended position estimated using a GHMM within a social forces based motion model yields a significant performance gain in comparison with the standard constant velocity-based models.