Navigation obstacle avoidance method combining unmanned aerial vehicle and unmanned ship
A technology of unmanned boats and unmanned aerial vehicles, which is applied in two-dimensional position/channel control and other directions, can solve the problems of low planning efficiency, low real-time performance, local optimum, etc., and achieve planning efficiency and planning real-time improvement, high intelligence sexual effect
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0023] In the process of autonomous navigation and obstacle avoidance of the unmanned boat, the main task is to realize the control of the unmanned boat through the control strategy generated by the decision-making algorithm based on the given starting point, end point and initial position of the obstacle. Successfully avoiding obstacles and reaching the end point under the guidance of a series of decision-making actions, realizing effective real-time intelligent algorithms is the fundamental to accomplish this task. Therefore:
[0024] (a) Calculate the decision-making action at the current moment while perceiving the state information of the UAV itself and the information of the surrounding environment and obstacles.
[0025] (b) A series of decision-making actions are generated to realize the overall process of navigation and obstacle avoidance.
[0026] (c) Each strategy in the set of decision control strategies is effective and easy to implement for the UAV.
[0027] For...
Embodiment 2
[0031] According to the problem description of UAV navigation and obstacle avoidance, the mathematical characteristics of this problem can be regarded as a Markov process in discrete time, and its components include state, action, transfer function, reward and so on. Usually, a state corresponds to an action or the probability of taking an action. When the action in the state is determined, the state after the transition can be known. To a certain extent, the good or bad of a certain state of the unmanned boat can be described by the evaluation value, so the return G is used. t It is used to represent the return that the unmanned boat state will have at a certain time t during the navigation and obstacle avoidance process:
[0032]
[0033] where G t represents the sum of discounts for immediate returns, and λ is the discount factor.
[0034] But in fact, when the whole decision-making process is not over, that is, the UAV has not reached the end point, has not collided w...
Embodiment 3
[0052] Because the traditional DQN generally overestimates the Q value of the decision-making action of the UAV, and the estimation error will accumulate as the number of actions increases. And the overestimation is not uniform, which leads to the overestimated Q value of a suboptimal UAV control action exceeding the Q value of the optimal control action, and the optimal strategy can never be found. Therefore, Dueling DQN is used here on the basis of DQN, and a dueling network is used to fit the Q value in UAV navigation and obstacle avoidance, but at the end of the network it is divided into two parts, that is, the state value function V(s ) represents the value of the static state environment itself, and the action advantage function A(a) represents the additional value brought by selecting an Action. The Q value is obtained by adding the state V value and the action A value. The purpose is to say that the state value is the same, but the advantages brought by each action ar...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More