Unmanned aerial vehicle autonomous flight control method based on offline reinforcement learning

Abstract

The invention discloses an unmanned aerial vehicle autonomous flight control method based on offline reinforcement learning, and the method comprises the following steps: (1) manually controlling an unmanned aerial vehicle to execute a flight task, collecting the flight data of the unmanned aerial vehicle in a real environment, and generating a data set; (2) designing a reward function according to the flight state and action based on the data set; (3) based on an offline reinforcement learning algorithm, training an autonomous flight control strategy only by using the data set; (4) in a real environment, controlling the unmanned aerial vehicle to execute a flight task by using an autonomous flight control strategy, monitoring and testing the performance of the control strategy by an unmanned aerial vehicle operator in real time, and collecting flight data; (5) adding the collected new flight data into the data set; and (6) iteratively executing the steps (2), (3), (4) and (5) until the autonomous flight control strategy can complete the flight task. According to the method, an autonomous flight control strategy with good generalization and robustness can be trained with low cost, and the method is suitable for a complex and changeable real environment.

Description

technical field [0001] The invention relates to an autonomous flight control method of an unmanned aerial vehicle based on off-line reinforcement learning, and belongs to the technical field of autonomous flight control of an unmanned aerial vehicle. Background technique [0002] Due to its simple structure, flexibility and mobility, UAV has been widely used, and has an irreplaceable position in the fields of aerial photography, surveying and mapping, logistics and navigation. UAV system control technology has developed rapidly in recent years, and autonomous flight technology is one of its core technologies. With the extensive application of UAVs in various scenarios, the flight environment faced by them is more complex and changeable, and the uncertainty is greatly increased. Traditional UAV autonomous flight methods can often achieve satisfactory results in some simple environments, but it is difficult to cope with flight tasks in highly dynamic environments. [0003] I...

AI Technical Summary

Problems solved by technology

[0005] Purpose of the invention: Aiming at the problem of high training cost of the reinforcement learning algorithm in the real environment, the present invention provides a UAV autonomous flight control method based on off-line reinforcement learning, which collects UAV flight data artificially, and does not require any connection with reality during the training process. Real-time interaction with the environment, using offline reinforcement learning algorithms, can train a generalized and robust autonomous flight control strategy at a very low cost, which is suitable for complex and changeable real environments

Images