.png){kind=icon}
.png){kind=icon}
Return
Safe and Adaptive UAV Path Planning via DRL and CBF Enhanced Dynamic Window Approach
.png){kind=icon}
PRE
AI
Z
吴
姚
L
DOI：10.1109/taes.2026.3713288.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The dynamic window approach (DWA) is a widely used local path planning algorithm for unmanned aerial vehicles (UAVs), offering online obstacle avoidance under dynamic constraints. However, the traditional DWA suffers from performance degradation due to conservative deceleration constraints and fixed evaluation weights. To address these limitations, this article proposes an improved DWA algorithm that integrates a discrete-time control barrier function and deep reinforcement learning to enhance adaptability. The discrete-time control barrier function formally defines safe regions and replaces the deceleration constraint and clearance function, enabling efficient obstacle avoidance without unnecessary deceleration. Meanwhile, the deep reinforcement learning is introduced to train a policy to adaptively adjust the DWA evaluation function weighting coefficients in response to the environment, balancing goal tracking, velocity, and obstacle avoidance. The proposed method is evaluated on the UAV within a 3-D environment populated with multiple static obstacles. Simulation results demonstrate that the proposed method achieves improved performance compared to the traditional DWA algorithm and offers greater safety assurance than recent approaches based on reinforcement learning.
Keywords:
Control barrier function (CBF)
deep reinforcement learning (DRL)
dynamic window approach (DWA)
local path planning
Journal
IF:
5.7
Papers:
651
Citations:
2.4W
Organization
Cited Papers