Crowd battle simulation method based on emotional infection and deep reinforcement learning

Abstract

The invention relates to the field of crowd simulation, reinforcement learning and statistics, and especially relates to a crowd battle simulation method based on emotional infection and deep reinforcement learning. The method comprises the following steps: extracting people number information of all agents; randomly endowing each agent with ID information, an initial position, an initial emotionvalue, an initial reward value and an initial action; obtaining an initial average physiological influence through initial action calculation; inputting the people number information, the ID information, the initial positions, the initial emotion values, the initial reward values, the initial actions and the initial average physiological influence into a neural network to obtain the action of theagent at the moment t; and sequentially carrying out iterative training for a plurality of times. Therefore, when a violent event occurs suddenly, strategy guidance can be quickly provided for the right side by the method, the reverse side is subdued as soon as possible, the situation that the scene is out of control is avoided, and high earnings are obtained, so social safety is maintained. Meanwhile, a simulation model also has great reference and learning values for real battle exercises.

Description

technical field [0001] The invention relates to the fields of crowd simulation, reinforcement learning and statistics, in particular to a crowd confrontation simulation method based on emotional infection and deep reinforcement learning. Background technique [0002] With the increase of the population, large-scale crowds gather, criminals gather to make troubles and even riots occur frequently, and the losses and casualties caused are shocking. Therefore, how to quell the war and maintain social security in a short period of time while minimizing losses is an important problem that needs to be solved urgently. At present, such problems are mainly solved through pre-manual exercises and the use of crowd simulation models. However, manual exercises will consume a lot of resources, and although exercises can improve combat skills and proficiency, sometimes it is impossible to quickly come up with a combat plan for some unexpected events. Therefore, we are more inclined to us...

AI Technical Summary

Problems solved by technology

Solve the problems of multi-dimensional explosion of the number of researchers and difficulty in convergence

Images