[Paper Review] Optimizing Collision Avoidance in Dense Airspace using Deep Reinforcement Learning

Sheng Li, Maxim Egorov+1 more|arXiv (Cornell University)|Dec 20, 2019

Autonomous Vehicle Technology and SafetyEngineering30 references34 citations

TL;DR

The paper develops and analyzes autonomous collision avoidance for dense airspace using deep reinforcement learning to correct an existing approach, achieving greater efficiency while preserving safety.

ABSTRACT

New methodologies will be needed to ensure the airspace remains safe and efficient as traffic densities rise to accommodate new unmanned operations. This paper explores how unmanned free-flight traffic may operate in dense airspace. We develop and analyze autonomous collision avoidance systems for aircraft operating in dense airspace where traditional collision avoidance systems fail. We propose a metric for quantifying the decision burden on a collision avoidance system as well as a metric for measuring the impact of the collision avoidance system on airspace. We use deep reinforcement learning to compute corrections for an existing collision avoidance approach to account for dense airspace. The results show that a corrected collision avoidance system can operate more efficiently than traditional methods in dense airspace while maintaining high levels of safety.

Motivation & Objective

• Motivate the need for advanced collision avoidance as airspace traffic density increases due to unmanned operations.
• Develop autonomous collision avoidance systems capable of operating in dense airspace where traditional methods struggle.
• Introduce metrics to quantify decision burden on the collision avoidance system and the system's impact on airspace.
• Apply deep reinforcement learning to compute corrections to an existing collision avoidance approach for dense airspace.
• Demonstrate that the corrected system can operate more efficiently while maintaining high safety standards.

Proposed method

• Use deep reinforcement learning to learn corrections to an existing collision avoidance approach.
• Introduce metrics for decision burden on the collision avoidance system and for its impact on airspace.
• Evaluate the corrected system in dense airspace scenarios and compare to traditional collision avoidance methods.
• Analyze performance in terms of efficiency and safety under high traffic densities.

Experimental results

Research questions

• RQ1 Can deep reinforcement learning improve the efficiency of collision avoidance in dense airspace without compromising safety?
• RQ2 What are effective metrics to quantify the decision burden on a collision avoidance system and its impact on airspace?
• RQ3 How does the RL-corrected collision avoidance approach compare to traditional methods under dense traffic conditions?
• RQ4 Is it feasible to derive corrections to existing collision avoidance strategies using deep learning?

Key findings

• A corrected collision avoidance system based on deep reinforcement learning can operate more efficiently than traditional methods in dense airspace.
• The RL-based corrections maintain high levels of safety alongside improved efficiency.
• The proposed metrics effectively capture decision burden and airspace impact, supporting the assessment of Dense Airspace collision avoidance performance.