MADDPG multi-agent reinforcement learning model-oriented visual analysis method

Abstract

The invention belongs to the technical field of information, and discloses an MADDPG multi-agent reinforcement learning model-oriented visual analysis method. The method comprises the steps of 1 selecting a cooperative game as a running environment of the MADDPG model, and defining a related parameter set; 2 training the MADDPG model, and storing and calculating important intermediate data; 3 designing a tag board, and marking an intelligent agent and a landmark; 4 designing a statistical view; 5 designing a commentator behavior view for evaluating the performance of the commentator obtained by model learning; and 6 designing an interactive view. The invention provides a novel visual analysis method which can support interactive analysis of the working process and the internal principle of the MADDPG model in the cooperation environment. According to the method, a plurality of collaborative views are designed, and the internal execution mechanism of the MADDPG model is disclosed from different angles.

Description

technical field [0001] The invention belongs to the field of information technology, in particular to a visual analysis method for a MADDPG multi-agent reinforcement learning model. Background technique [0002] Deep reinforcement learning is a very hot research field today, and it has been used to solve various challenging application problems such as autonomous driving, traffic control, and robotic system control. Despite the superior performance of deep reinforcement learning in these applications, researchers still know little about their intrinsic execution mechanism. In recent years, researchers have proposed various visual analysis methods to improve the interpretability of deep reinforcement learning models. For Q-Network (DQN), a visual analysis system DQNViz is designed to reveal the agent's experience space from different levels. For the dueling DQN and Asynchronous Advantage Actor-Critic models, generate saliency maps to show which parts of the input image the ...

AI Technical Summary

Problems solved by technology

Compared with single-agent deep reinforcement learning, analyzing multi-agent deep reinforcement learning models is more challenging, mainly because: 1) The increase in the number of agents leads to an exponential growth of the state space, how to visualize the experience generated by multiple agents spaces and reveal potential connections between them? 2) Multiple agents are constantly interacting with different environmental objects (landmarks), how to intuitively visualize the interaction process over time? Existing research lacks interpretability research on multi-agent deep reinforcement learning models

Images