A method and system for autonomous flight control of unmanned helicopter

Abstract

The invention discloses an autonomous flight control method and system of an unmanned helicopter, belonging to the technical field of unmanned aerial vehicles. * and the state s of the unmanned helicopter as the input value of the prediction model, the prediction model outputs the predicted state s of the unmanned helicopter at the next moment next ; will expect output x d and the state of the unmanned helicopter at the next moment next As the input value of the control model, the control model outputs the optimal control quantity u for autonomous flight of the unmanned helicopter * And upload it to the unmanned helicopter to control the autonomous flight of the unmanned helicopter. The prediction model and the control model are convex time convolutional neural network models with different coefficients. Compared with the general deep neural network, the training speed of this method is faster, the model accuracy is high, and the size of the receptive field can be adjusted arbitrarily.

Description

technical field [0001] The invention relates to the technical field of unmanned aerial vehicles, in particular to an autonomous flight control method and system of an unmanned helicopter. Background technique [0002] Unmanned helicopters not only have outstanding flight characteristics such as side flight, rear flight, vertical take-off and landing, hovering, and obstacle avoidance maneuvers, but also have the advantages of small size, light weight, easy concealment, and low price, so they are widely used in both military and civil fields. with broadly application foreground. Unmanned helicopter is a complex nonlinear system that is statically unstable, strongly coupled, underactuated, and high-order time-varying, which brings great challenges to the system modeling and flight control of unmanned helicopters. [0003] The autonomous flight technology of unmanned helicopters is a cutting-edge technology for national defense and has been facing the technological blockade of ...

AI Technical Summary

Problems solved by technology

At present, there are mainly two types of methods commonly used. The first type of method uses linearization, piecewise linearization, or Gaussian model to construct a simple system model. The model constructed by this type of method is convenient for the realization of subsequent flight control and can be used for theoretical analysis. and proof, but as the difficulty of flight control tasks increases, the model representation ability is insufficient and cannot be used for complex flight control tasks; the second type of method constructs a system model through a complex machine learning model (deep neural network, SVM, etc.), such The method can accurately capture the dynamic relationship of the system, and the model has strong expressive ability, but this kind of method is very easy to fall into the local minimum, which brings great difficulties to the subsequent flight control

Images