SwordRiding: A Unified Navigation Framework for Quadrotors in Unknown Complex Environments via Online Guiding Vector Fields

Xuchen Liu, Ruocheng Li, Bin Xin, Weijia Yao, Qigeng Duan, Jinqiang Cui, Ben M. Chen, Jie ChenarXiv 2025 · DOI · arXiv

Abstract

Traditional quadrotor navigation often relies on open-loop trajectory tracking, which struggles with unpredictable external disturbances. We introduce SwordRiding, a navigation framework that implements a "closed-loop" paradigm by converting discretized B-spline paths into Online Guiding Vector Fields (GVFs). This creates a continuous guidance "flow" that pulls the drone back to its safe trajectory regardless of its position in space. Our method decouples geometric path planning from algebraic constraints, enabling real-time, robust flight in unknown, cluttered environments. Experimental results in both simulation and real-world scenarios demonstrate a 92% success rate under heavy disturbances, significantly outperforming existing state-of-the-art planners.

Introduction: The Open-Loop Challenge

Most modern drone navigation systems follow an open-loop planning paradigm. A high-level planner generates a time-indexed trajectory, and a low-level controller attempts to track it. While efficient, this approach is fragile: if a drone is pushed off-course by a gust of wind or a physical collision, the high-level plan remains "unaware," often leading to jerky corrections or total navigation failure.

Comparison of (a) traditional open-loop trajectory tracking and (b) the proposed closed-loop guiding vector field architecture.

Existing solutions like Guiding Vector Fields (GVF) offer a "closed-loop" alternative where the robot is guided by a vector field pointing toward the path from any position. However, traditional GVF methods require paths defined by simple mathematical functions (like circles or lines), making them difficult to apply to the complex, real-time paths generated in unknown environments.

The SwordRiding Framework

SwordRiding bridges the gap between flexible path planning and robust GVF guidance through a multi-stage pipeline:

  1. Perception & Mapping: Using LiDAR-based SLAM (FAST-LIO2), the system builds a Euclidean Signed Distance Field (ESDF), providing real-time distance-to-obstacle data.
  2. Trajectory Optimization: A global planner finds a collision-free path, which is then smoothed via uniform B-splines to ensure physical feasibility.
  3. Incremental GVF Construction: This core innovation discretizes the B-spline path and uses the ESDF to calculate inward normal vectors (pointing to the path) and tangential vectors (pointing along the path) via local polynomial fitting.
  4. Closed-Loop Guidance: These vectors form a field that automatically generates commands to steer the drone back to its path if displaced, without needing a manual reset or replanning.
The SwordRiding pipeline: from obstacle observation and trajectory discretization to the generation of a continuous guiding vector field.

Key Insights

  • Decoupling Geometry from Algebra: By using distance transforms on a discretized path, SwordRiding turns arbitrary, jagged paths into smooth guiding fields without needing explicit mathematical equations.
  • Inherent Disturbance Rejection: Navigation commands are derived from the drone's current position relative to the field, baking "correction" directly into the high-level guidance.
  • Edge Efficiency: The entire pipeline—from mapping to vector field generation—runs at 20Hz on embedded hardware (NVIDIA Orin NX), proving that complex closed-loop planning is viable for real-time flight.

Experimental Results

The framework was evaluated against Fast Planner and Ego Planner in both simulation and real-world tests.

  • Robustness to Disturbance: In simulations with external force perturbations (simulating 1.5 m/s² acceleration), SwordRiding maintained a 92% success rate, while competitors dropped to ~23% as they could not recover from being pushed off-trajectory.
  • Real-World Recovery: In "Manual Dragging" tests, researchers physically moved the drone mid-flight. Upon release, the GVF immediately re-oriented and guided the drone back to its original path without manual intervention.
  • Performance: Average planning time was approximately 9.9 ms, ensuring high responsiveness to environmental changes.

Conclusion

SwordRiding demonstrates that closed-loop vector field navigation is not only possible but highly effective for quadrotors in complex environments. While current work focuses on robustness, future iterations will aim to improve tracking precision at high speeds and adapt the framework for underwater robotics, where unpredictable currents make robust guidance even more critical.

Summary updated

Abstract

Although quadrotor navigation has achieved high performance in trajectory planning and control, real-time adaptability in unknown complex environments remains a core challenge. This difficulty mainly arises because most existing planning frameworks operate in an open-loop manner, making it hard to cope with environmental uncertainties such as wind disturbances or external perturbations. This paper presents a unified real-time navigation framework for quadrotors in unknown complex environments, based on the online construction of guiding vector fields (GVFs) from discrete reference path points. In the framework, onboard perception modules build a Euclidean Signed Distance Field (ESDF) representation of the environment, which enables obstacle awareness and path distance evaluation. The system first generates discrete, collision-free path points using a global planner, and then parameterizes them via uniform B-splines to produce a smooth and physically feasible reference trajectory. An adaptive GVF is then synthesized from the ESDF and the optimized B-spline trajectory. Unlike conventional approaches, the method adopts a closed-loop navigation paradigm, which significantly enhances robustness under external disturbances. Compared with conventional GVF methods, the proposed approach directly accommodates discretized paths and maintains compatibility with standard planning algorithms. Extensive simulations and real-world experiments demonstrate improved robustness against external disturbances and superior real-time performance.
BibTeX
@misc{liu2025swordriding,
  author = {Xuchen Liu and Ruocheng Li and Bin Xin and Weijia Yao and Qigeng Duan and Jinqiang Cui and Ben M. Chen and Jie Chen},
  title = {{SwordRiding: A Unified Navigation Framework for Quadrotors in Unknown Complex Environments via Online Guiding Vector Fields}},
  year = {2025},
  eprint = {2511.22043},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2511.22043},
}
Figure