Search papers, labs, and topics across Lattice.
This paper presents DPNet, a lightweight neural network designed to predict and avoid dead ends in vision-based UAV navigation using RGB-D inputs. By efficiently estimating the relative distance and bearing of potential dead ends, the method enhances trajectory planning by pruning a compact trajectory library, allowing for proactive navigation while ensuring smooth flight. The approach shows significant improvements in success rate, flight time, and trajectory length in both simulations and real-world scenarios, achieving high-frequency replanning at 50 Hz without the need for manual data annotation or fine-tuning.
UAVs can now navigate complex environments more effectively by predicting dead ends in real-time, leading to smoother and more efficient flight paths.
Vision-based Unmanned Aerial Vehicles (UAVs) often suffer from navigation failures in dead ends due to limited sensing accuracy and range. To address this challenge, this paper proposes a systematic solution for efficient dead-end prediction and avoidance. The proposed method introduces a lightweight neural network to predict the relative distance and bearing of potential dead ends within the current field of view using RGB-D inputs. These predictions prune a predefined, compact trajectory library, enabling the planner to proactively avoid dead ends while maintaining navigational smoothness. Notably, our approach transfers across real-world scenarios without manual annotation or fine-tuning on real-world data. The system achieves high-frequency replanning at 50 Hz onboard. Extensive simulation benchmarks demonstrate superior performance in success rate, flight time, and trajectory length, and real-world experiments further validate its effectiveness in complex scenarios.
