IntentNet: Learning to Predict Intention from Raw Sensor Data

TL;DR

This paper introduces IntentNet, a unified model that predicts traffic participants' intentions using LiDAR data and environmental maps, improving accuracy and efficiency for autonomous vehicle planning.

Contribution

IntentNet is the first one-stage model combining intent detection and trajectory forecasting from raw sensor data for self-driving cars.

Findings

Outperforms separate models in accuracy

Reduces computation time

Enhances reaction speed in autonomous driving

Abstract

In order to plan a safe maneuver, self-driving vehicles need to understand the intent of other traffic participants. We define intent as a combination of discrete high-level behaviors as well as continuous trajectories describing future motion. In this paper, we develop a one-stage detector and forecaster that exploits both 3D point clouds produced by a LiDAR sensor as well as dynamic maps of the environment. Our multi-task model achieves better accuracy than the respective separate modules while saving computation, which is critical to reducing reaction time in self-driving applications.