.png){kind=icon}
.png){kind=icon}
Return
Control-oriented UAV highly feasible trajectory planning: A deep learning method
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ast.2020.106435.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The highly feasible trajectory planning of unmanned aerial vehicle (UAV) is very important in some tasks but has not yet attracted sufficient study attention. Most current studies use simplified UAV model with some state constraints to plan the trajectory, but the feasibility is reduced, because the simplified model is very different from the actual UAV system, so that the tracking characteristics of UAV cannot be fully considered. In this paper, a novel control-oriented UAV highly feasible trajectory planning method is proposed. First, a UAV closed-loop model prediction method, which is the combination of a low-level controller and a UAV 6 DOF nonlinear model, is adopted in the trajectory planning phase to predict the flight trajectory. This complicated model is very similar to the actual UAV system because it comprehensively considers the controller performance and the detailed UAV model, but it also has poor efficiency. Therefore, a trajectory-mapping network (TMN) is proposed using a deep learning approach to improve the planning efficiency. Furthermore, a novel time-series convolutional neural network (TSCNN) is proposed for the TMN to further improve its computation speed and prediction accuracy. Finally, the flight trajectory predicted by the TMN is used to evaluate the planning cost. In this way, the planned trajectory will be highly feasible. The effectiveness of the proposed method is demonstrated by simulations. (C) 2020 Elsevier Masson SAS. All rights reserved.
Keywords:
Highly feasible trajectory planning
Trajectory-mapping network
Deep learning
Unmanned aerial vehicle
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
5.8
Papers:
1.0W
Citations:
3.0W