Dynamically Local-Enhancement Planner for Large-Scale Autonomous Driving

TL;DR

This paper introduces the DLE Planner, a method that dynamically enhances a basic autonomous driving policy with local data using GNNs, improving scalability and safety without enlarging the model.

Contribution

The paper presents a novel dynamically local-enhancement approach that improves large-scale autonomous driving systems by integrating local features without modifying the core planner.

Findings

Outperforms baseline in safety and reward

Maintains a lighter model scale

Effective in multiple driving scenarios

Abstract

Current autonomous vehicles operate primarily within limited regions, but there is increasing demand for broader applications. However, as models scale, their limited capacity becomes a significant challenge for adapting to novel scenarios. It is increasingly difficult to improve models for new situations using a single monolithic model. To address this issue, we introduce the concept of dynamically enhancing a basic driving planner with local driving data, without permanently modifying the planner itself. This approach, termed the Dynamically Local-Enhancement (DLE) Planner, aims to improve the scalability of autonomous driving systems without significantly expanding the planner's size. Our approach introduces a position-varying Markov Decision Process formulation coupled with a graph neural network that extracts region-specific driving features from local observation data. The learned features describe the local behavior of the surrounding objects, which is then leveraged to enhance a basic reinforcement learning-based policy. We evaluated our approach in multiple scenarios and compared it with a one-for-all driving model. The results show that our method outperforms the baseline policy in both safety (collision rate) and average reward, while maintaining a lighter scale. This approach has the potential to benefit large-scale autonomous vehicles without the need for largely expanding on-device driving models.