TridentNetV2: Lightweight Graphical Global Plan Representations for Dynamic Trajectory Generation

TL;DR

This paper introduces a lightweight, map-independent framework for autonomous vehicle trajectory generation that models feasible paths in real-time using minimal scene information, achieving high accuracy and efficiency.

Contribution

It proposes a novel approach that replaces HD maps with a lightweight scene representation for real-time trajectory planning in autonomous navigation.

Findings

Achieves low error rates across urban navigation datasets.

Maintains real-time performance with reduced network complexity.

Effectively models feasible trajectories without relying on detailed HD maps.

Abstract

We present a framework for dynamic trajectory generation for autonomous navigation, which does not rely on HD maps as the underlying representation. High Definition (HD) maps have become a key component in most autonomous driving frameworks, which include complete road network information annotated at a centimeter-level that include traversable waypoints, lane information, and traffic signals. Instead, the presented approach models the distributions of feasible ego-centric trajectories in real-time given a nominal graph-based global plan and a lightweight scene representation. By embedding contextual information, such as crosswalks, stop signs, and traffic signals, our approach achieves low errors across multiple urban navigation datasets that include diverse intersection maneuvers, while maintaining real-time performance and reducing network complexity. Underlying datasets introduced are available online.