Stratospheric airship height control method and system
A stratospheric airship, altitude control technology, applied in instrumentation, geometric CAD, design optimization/simulation, etc., can solve problems such as difficult self-adaptation, difficult parameter selection, and large randomness
Image
Examples
Embodiment 1
[0039]In this embodiment, aiming at the altitude control problem of the stratospheric airship, a Markov decision process model consisting of determining the state transition probability based on the airship speed distribution and the return value space based on the obstacle avoidance idea is proposed, and a highly intelligent control method using the Q-learning algorithm, Taking the speed distribution as the basis for the selection of action probability in the Q-learning learning process is one of the key elements of the algorithm. In order to realize the interaction between the external environment and the agent, the speed needs to be corrected by feedback, and the real-time height difference is used as the ascent speed distribution update. The basis to ensure the reasonable validity of the state transition probability. The reward value space based on the idea of ββobstacle avoidance is the main basis for judging the optimal action sequence. Through the expected height change ...
Embodiment 2
[0079] This embodiment is used to provide a kind of stratospheric airship height control system, utilizes the airship height control method as described in embodiment 1 to work, and described control system comprises:
[0080] The Markov decision process model acquisition module is used to establish the Markov decision process model for airship height control; the Markov decision process model includes state space, action space, return value space, state transition probability and target optimization function;
[0081] The Markov decision process model acquisition module includes a return value space design submodule; the return value space design submodule includes:
[0082] a grid space acquiring unit, configured to acquire a pre-planned expected height change curve; and perform rasterization on the surrounding space of the expected height change curve to obtain a grid space;
[0083] A reward value calculation unit, configured to select an action in the action space for ea...
PUM
Login to View More Abstract
Description
Claims
Application Information
- IPC
- G06F30/15; G06F30/20; G06N7/00; G06F111/08
- CPC
- G06F30/15; G06F30/20; G06F2111/08; G06N7/01; Y02T90/00
- Inventors
- ζ¨εΈη₯₯; ζ¨ζδΌ