Bionic robotic fish cluster navigation simulation method based on deep reinforcement learning technology

Abstract

The invention provides a bionic robotic fish cluster navigation simulation method based on the deep reinforcement learning technology, and relates to the technical field of multi-agent path navigation planning. The method comprises the following steps: firstly, constructing a 3D fish swarm environment model, and then constructing an agent model of a bionic robotic fish cluster in the 3D fish swarm environment; the intelligent agent model comprises a perception model, a movement model and a decision-making model; constructing a reward function of the fish swarm cluster, and introducing a curiosity mechanism into the reward function; constructing a distributed training framework of an agent model based on a curiosity mechanism and a PPO2 algorithm, and enabling an agent to obtain a behavior strategy in a learning mode; and finally, training an agent model based on the constructed distributed training framework to realize navigation simulation of the bionic robotic fish cluster. According to the method, a virtual fish school can learn reasonable fish school behaviors in a 3D environment, and the method is applied to bionic robot fish school navigation behaviors in the real world.

Description

technical field [0001] The invention relates to the technical field of multi-agent path navigation planning, in particular to a bionic robot fish swarm navigation simulation method based on deep reinforcement learning technology. Background technique [0002] The flocking behavior of fish is a typical self-organization phenomenon. In the process of swimming, fish will naturally gather in groups to ensure their own survival, and display complex group behaviors. The swimming of each fish only needs to follow two basic rules: follow the fish next to it; keep moving. If only based on these two simple rules to simulate the behavior of natural fish swarms and realize bionic robot fish swarm navigation, it is difficult to realize most artificial swarm systems at present. [0003] A common method used to simulate fish swarm behavior to realize bionic robot fish swarm navigation is Artificial Fish Swarm Algorithm (AFSA). Artificial fish swarm algorithm is an optimization algorithm...

AI Technical Summary

Problems solved by technology

[0005] Artificial fish swarm algorithm (AFSA) has problems such as low convergence accuracy, easy to fall into local optimum, and slow convergence speed in the later stage.

The algorithm is very sensitive to various hyperparameters and is easily affected by step size, population size and crowding factor, so it has great limitations

[0006] In order to simplify the training, the current method of using deep reinforcement learning to simulate the flocking behavior of fish schools is mostly a simple 2D environment, with small action space and state space, which cannot truly reflect the flocking behavior of fish flocks in nature.

Images