- Abstract: Intention recognition of pedestrians is crucial to safe and reliable working of autonomous vehicles, when serving as, for instance, indoor service robots or self-driving cars in busy urban scenes. Previously, Chen et al.  combined Markovian-based and clustering-based approaches to learn motion primitives and subsequently predict pedestrian trajectories by modeling the transition between learned primitives as a Gaussian Process (GP). This work further develops their approach by incorporating semantic features from the environment (relative distance to curbside and status of pedestrian traffic lights) for more confident prediction of pedestrian trajectories at intersections. Adding the environmental context, when available, not only makes prediction more robust but can also provide increased flexibility of prediction in new environments. We test our algorithm on real data. The results show 26% improvement in prediction accuracy as compared to previous work, on incorporation of new features.
- Keywords: Pedestrian intention, Self-driving cars, Gaussian Process