Agent Prioritization for Autonomous Navigation

TL;DR

This paper introduces a real-time agent ranking system for autonomous vehicles that prioritizes agents based on their impact on navigation decisions, utilizing learned and engineered features for improved understanding.

Contribution

It presents a novel method to automatically generate training data for agent importance ranking and combines learned and engineered features in a neural network model.

Findings

The system effectively ranks agents in complex driving scenarios.

Combining learned and engineered features improves ranking accuracy.

Real-world tests show enhanced understanding of driving situations.

Abstract

In autonomous navigation, a planning system reasons about other agents to plan a safe and plausible trajectory. Before planning starts, agents are typically processed with computationally intensive models for recognition, tracking, motion estimation and prediction. With limited computational resources and a large number of agents to process in real time, it becomes important to efficiently rank agents according to their impact on the decision making process. This allows spending more time processing the most important agents. We propose a system to rank agents around an autonomous vehicle (AV) in real time. We automatically generate a ranking data set by running the planner in simulation on real-world logged data, where we can afford to run more accurate and expensive models on all the agents. The causes of various planner actions are logged and used for assigning ground truth importance scores. The generated data set can be used to learn ranking models. In particular, we show the utility of combining learned features, via a convolutional neural network, with engineered features designed to capture domain knowledge. We show the benefits of various design choices experimentally. When tested on real AVs, our system demonstrates the capability of understanding complex driving situations.