FlowPilot: Real-Time World-Action Modeling for Agile UAV Navigation
FlowPilot: Real-Time World-Action Modeling for Agile UAV Navigation
This research presents FlowPilot, the first world-action model designed for real-time onboard UAV navigation using depth data.
The paper introduces FlowPilot, a world-action model for real-time UAV navigation that jointly denoises depth and trajectories using flow matching; it reports closed-loop simulation performance under clutter up to 10 m/s and physical deployment on a quadrotor with 10.2 ms inference latency on a Jetson Orin NX.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
peak speed
10
View original evidence
t achieves a mean inference latency of 10.2 ms on a Jetson Orin NX and reaches a peak speed of 10 m/s during outdoor flight using only onboard sensing and computation.
Open source S5inference latency
10.2
View original evidence
t achieves a mean inference latency of 10.2 ms on a Jetson Orin NX and reaches a peak speed of 10 m/s during outdoor flight using only onboard sensing and computation.
Open source S5
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
e control points, yielding C^2-continuous references with closed-form velocity, acceleration and jerk. FlowPilot is trained on a three-level depth pyramid spanning high-throughput simulation, photorealistic simulation, and real onboard data. In closed-loop simulation, it outperforms learning- and optimization-based baselines under increasing clutter and commanded speeds up to 10 m/s. On a physical quadrotor, FlowPilo
Open source S4
t achieves a mean inference latency of 10.2 ms on a Jetson Orin NX and reaches a peak speed of 10 m/s during outdoor flight using only onboard sensing and computation.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org