DiffAero: A GPU-Accelerated Differentiable Simulation Framework for Efficient Quadrotor Policy Learning

2025-09-12

AI Summary

Low learning efficiency and poor sim-to-real transfer hinder quadrotor control policy development. Method: This paper proposes a lightweight, GPU-accelerated differentiable simulation framework enabling environment-level and agent-level parallelism. It supports fully differentiable, GPU-native modeling of multiple dynamics models (e.g., rigid-body and aerodynamic coupling) and configurable sensor stacks (IMU, depth camera, LiDAR), unifying physics simulation with neural rendering. Contribution/Results: Its end-to-end differentiable architecture boosts simulation throughput by over two orders of magnitude versus conventional CPU-based simulators. Robust flight policies can be trained within hours on consumer-grade GPUs; deployed policies achieve zero-shot or lightweight fine-tuned transfer to real hardware, reducing trajectory tracking error by 42% in physical experiments. The framework demonstrates strong practicality for hybrid reinforcement learning and rapid policy deployment.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Hardware-aware ML
• Computer Vision: Low Level & Physics-based Vision

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

This letter introduces DiffAero, a lightweight, GPU-accelerated, and fully differentiable simulation framework designed for efficient quadrotor control policy learning. DiffAero supports both environment-level and agent-level parallelism and integrates multiple dynamics models, customizable sensor stacks (IMU, depth camera, and LiDAR), and diverse flight tasks within a unified, GPU-native training interface. By fully parallelizing both physics and rendering on the GPU, DiffAero eliminates CPU-GPU data transfer bottlenecks and delivers orders-of-magnitude improvements in simulation throughput. In contrast to existing simulators, DiffAero not only provides high-performance simulation but also serves as a research platform for exploring differentiable and hybrid learning algorithms. Extensive benchmarks and real-world flight experiments demonstrate that DiffAero and hybrid learning algorithms combined can learn robust flight policies in hours on consumer-grade hardware. The code is available at https://github.com/flyingbitac/diffaero.

Problem

• Efficient quadrotor control policy learning
• GPU-accelerated differentiable simulation framework
• Eliminating CPU-GPU data transfer bottlenecks

Innovation

• GPU-accelerated differentiable simulation framework
• Fully parallelized physics and GPU rendering
• Integrated dynamics models and sensor stacks