Pedestrian Prediction by Planning using Deep Neural Networks

TL;DR

This paper presents a neural network-based system that predicts pedestrian destinations and trajectories by emulating their motion planning, trained with inverse reinforcement learning, to improve autonomous vehicle collision avoidance.

Contribution

It introduces a novel approach combining destination inference and motion planning within a single neural network trained via inverse reinforcement learning.

Findings

Accurately predicts pedestrian destinations and trajectories

Demonstrates effectiveness on real-world data

Enhances collision avoidance capabilities

Abstract

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.