A Cooperative Game Path Planning Method Based on Neural Network and Artificial Potential Field

Abstract

The invention proposes a collaborative game path planning method based on a neural network and an artificial potential field, which belongs to the field of intelligent body dynamic path planning. This method first obtains the optimized path of the agent in the offline stage to construct a training sample set, and trains the BP neural network module to obtain the trained BP neural network module; in the online stage, obtains the current position and environment of each agent The information is input into the trained BP neural network module, and the network module outputs the repulsion gain at that moment, and then calculates the gravitational force of the target point on the agent at this moment, the repulsion force of the threat area on the agent, and calculates the resultant force. The agent moves according to the resultant force, and updates the repulsion gain according to the new position and environment information at the next moment until the path planning reaches the end condition. The present invention can better solve the path planning problem in the collaborative game scene, and can better adapt to the dynamic path planning under the movement of objects and obstacles.

Description

technical field [0001] The invention belongs to the field of intelligent body dynamic path planning, in particular to a collaborative game path planning method based on neural network and artificial potential field. Background technique [0002] The collaborative game path planning based on neural network and artificial potential field refers to the path planning that multi-agents use the method of combining the improved neural network and artificial potential field to collaboratively aim at reaching the target area and avoiding the obstacle area. , where the game refers to the real-time adversarial feedback of the target area and the obstacle area for the multi-agent movement. The path planning problem of cooperative game is an important branch of the dynamic path planning problem of particles in two-dimensional space. The research of this problem is of great significance to practical applications such as robot round-up and robot football game. [0003] Existing methods su...

AI Technical Summary

Problems solved by technology

[0019] The main disadvantages of this method are: poor adaptability to dynamic scenes, the potential field function cannot consider the movement of targets and obstacles, and path planning with prediction and pre-quantity cannot be performed

[0039] The main disadvantage of this method is that it lacks adaptability to cooperative path planning in game scenarios, and the synergy and game nature of its path planning are mainly reflected in the formulation of fuzzy rules. However, fuzzy rules need to be manually formulated, and the quality is difficult to guarantee. Difficulty, it is not as good as learning from samples; on the other hand, this method improves the formula in the classical artificial potential field method, but the improvement is directly for the attraction and repulsion functions, not for the potential field function, according to "by The principle of "field to force", the theoretical basis of this method to skip the field function and directly improve the force function needs to be further demonstrated

Images