Trajectory Generation by Sparse Demonstration Learning and Minimum Snap-based Optimization

In this paper, dynamic time warping function is used to establish an optimal control system for four-rotor unmanned aerial vehicle (UAV) to learn how to optimize trajectory planning from sparse demonstration. By continuous Pontryagin Differentiable Programming, UAV learns the objective function base...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2024 29th International Conference on Automation and Computing (ICAC) s. 1 - 6
Hlavní autoři: Xu, Taoying, She, Haoping, Si, Weiyong, Li, Chuanjun
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 28.08.2024
Témata:
Autonomous aerial vehicles
continuous Pontryagin Differentiable Programming
dynamic time warping
Heuristic algorithms
Linear programming
minimum snap
Optimal control
Planning
Programming
Resource management
sparse demonstration learning
Trajectory
Trajectory planning
Vehicle dynamics
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this paper, dynamic time warping function is used to establish an optimal control system for four-rotor unmanned aerial vehicle (UAV) to learn how to optimize trajectory planning from sparse demonstration. By continuous Pontryagin Differentiable Programming, UAV learns the objective function based on sparse waypoints demonstration. However, due to the small sample data of sparse demonstration learning, there is a problem of low precision, and Pontryagin's Minimum Principle itself has the limitation of easily falling into the local optimal solution. So, this paper adopts the Minimum Snap trajectory algorithm that meets the dynamic constraints of the agent to generate a planned trajectory, to weighted combination with learning trajectory solved based on continuous Pontryagin Differentiable Programming, and the resulting optimized trajectory has the advantages of small demonstration learning difference loss, reasonable time allocation and reasonable planning, so that UAV can have certain generalization capability and optimize a reasonable trajectory with less energy loss. Finally, the feasibility of the proposed method is verified by the simulation experiment of the four-rotor UAV.
DOI:10.1109/ICAC61394.2024.10718821