Pedestrian Prediction by Planning using Deep Neural Networks
This work addresses pedestrian prediction for autonomous vehicles, which is crucial for collision avoidance, but it appears incremental as it builds on existing planning and neural network approaches.
The paper tackles pedestrian trajectory prediction by emulating their motion planning, using a monolithic neural network trained via inverse reinforcement learning to infer destinations and trajectories. Experimental results on real-world data demonstrate accurate predictions of both destinations and trajectories.
Accurate traffic participant prediction is the prerequisite for collision avoidance of autonomous vehicles. In this work, we predict pedestrians by emulating their own motion planning. From online observations, we infer a mixture density function for possible destinations. We use this result as the goal states of a planning stage that performs motion prediction based on common behavior patterns. The entire system is modeled as one monolithic neural network and trained via inverse reinforcement learning. Experimental validation on real world data shows the system's ability to predict both, destinations and trajectories accurately.