Robust Global-Local Behavior Arbitration via Continuous Command Fusion Under LiDAR Errors
Summary: arXiv:2603.27273v1 Announce Type: cross
Abstract
Modular autonomous driving systems must coordinate global progress objectives with local safety-driven reactions under imperfect sensing and strict real-time constraints. This paper presents a ROS2-native arbitration module that continuously fuses the outputs of two unchanged and interpretable controllers: a global reference-tracking controller based on Pure Pursuit and a reactive LiDAR-based Gap Follow controller.
At each control step, both controllers propose Ackermann commands, and a PPO-trained policy predicts a continuous gate from a compact feature observation to produce a single fused drive command, augmented with practical safety checks. For comparison under identical ROS topic inputs and control rate, we implement a lightweight sampling-based predictive baseline.
Introduction
In the field of autonomous driving, ensuring safety while maintaining efficiency poses significant challenges, particularly in environments where sensor data is imperfect. The integration of robust algorithms capable of balancing global navigation with immediate safety responses is critical.
Methodology
This study introduces a novel arbitration module designed for ROS2, combining two distinct controllers: the global reference-tracking controller utilizing Pure Pursuit methodology, and a reactive Gap Follow controller that leverages LiDAR data for immediate local responses.
Controller Design
- Global Reference-Tracking Controller: Utilizes Pure Pursuit for maintaining a set trajectory.
- Reactive LiDAR-Based Controller: Identifies gaps in the environment to facilitate safe navigation.
Both controllers generate Ackermann commands at every control step. The integration of a Proximal Policy Optimization (PPO)-trained policy allows for the prediction of a continuous gating mechanism, effectively synthesizing commands into a single coherent output while implementing safety checks to mitigate risks.
Evaluation and Results
To assess the robustness of the proposed system, a series of experiments were conducted using a ROS2 impairment protocol. This protocol injected various levels of LiDAR noise, delays, and dropout conditions, simulating real-world scenarios that autonomous vehicles may encounter.
Testing Conditions
- LiDAR Noise Injection
- Delay Simulation
- Dropout Conditions
- Forward-Cone False Short-Range Outlier Sweeps
The experimental framework included a repeatable close-proximity passing scenario, where we measured both safe success and failure rates. Furthermore, we documented the end-to-end controller runtime at each step, particularly as the sensing stress intensified.
Conclusion
This study emphasizes a command-level robustness evaluation within a modular ROS2 architecture, providing insights into the efficacy of continuous command fusion in the face of LiDAR inaccuracies. The findings contribute to the ongoing discourse on enhancing safety and reliability in autonomous driving systems, underscoring the importance of effective global-local coordination.