Curriculum Reinforcement Learning for Quadrotor Racing with Random Obstacles

Autonomous drone racing has attracted increasing interest as a research topic for exploring the limits of agile flight. However, existing studies primarily focus on obstacle-free racetracks, while the perception and dynamic challenges introduced by obstacles remain underexplored, often resulting in low success rates and limited robustness in real-world flight. To this end, we propose a novel vision-based curriculum reinforcement learning framework for training a robust controller capable of addressing unseen obstacles in drone racing. We combine multi-stage cu rriculum learning, domain randomization, and a multi-scene updating strategy to address the conflicting challenges of obstacle avoidance and gate traversal. Our end-to-end control policy is implemented as a single network, allowing high-speed flight of quadrotors in environments with variable obstacles. Both hardware-in-the-loop and real-world experiments demonstrate that our method achieves faster lap times and higher success rates than existing approaches, effectively advancing drone racing in obstacle-rich environments. The video and code are available at: https://github.com/SJTU-ViSYS-team/CRL-Drone-Racing.

• Publication year

  2026

• Venue

  Unknown venue

• Publication date

  2026-02-27

• Fields of study

  Computer Science, Engineering

• Identifiers
  arXiv 2602.24030
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Seeing Through Pixel Motion: Learning Obstacle Avoidance From Optical Flow With One Camera

  2024cited by this paper
• Learning Speed Adaptation for Flight in Clutter

  2024cited by this paper
• MAVRL: Learn to Fly in Cluttered Environments With Varying Speed

  2024cited by this paper
• Multi-Task Reinforcement Learning for Quadrotors

  2024cited by this paper
• Learning Generalizable Policy for Obstacle-Aware Autonomous Drone Racing

  2024influential reference
• VisFly: An Efficient and Versatile Simulator for Training Vision-Based Flight

  2024cited by this paper
• Learning vision-based agile flight via differentiable physics

  2024cited by this paper
• Demonstrating Agile Flight from Pixels without State Estimation

  2024cited by this paper
• Champion-level drone racing using deep reinforcement learning

  2023cited by this paper
• Autonomous Drone Racing: A Survey

  2023cited by this paper
• Reaching the limit in autonomous racing: Optimal control versus reinforcement learning

  2023cited by this paper
• Learning Perception-Aware Agile Flight in Cluttered Environments

  2022cited by this paper
• Minimum-Time Quadrotor Waypoint Flight in Cluttered Environments

  2022cited by this paper
• Learning Minimum-Time Flight in Cluttered Environments

  2022influential reference
• Model Predictive Contouring Control for Time-Optimal Quadrotor Flight

  2021cited by this paper
• Fast-Racing: An Open-Source Strong Baseline for SE(3) Planning in Autonomous Drone Racing

  2021cited by this paper
• Learning high-speed flight in the wild

  2021cited by this paper
• Habitat: A Platform for Embodied AI Research

  2019cited by this paper
• Teach-Repeat-Replan: A Complete and Robust System for Aggressive Flight in Complex Environments

  2019cited by this paper
• PAMPC: Perception-Aware Model Predictive Control for Quadrotors

  2018cited by this paper
• Proximal Policy Optimization Algorithms

  2017cited by this paper
• Whole-Body Nonlinear Model Predictive Control Through Contacts for Quadrupeds

  2017cited by this paper
• Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

  2014cited by this paper
• Geometric tracking control of a quadrotor UAV on SE(3)

  2010cited by this paper

• No citing papers are available for this paper.