多智能体任务分配

自己的小项目围绕强化学习展开：先用 Gym 搭建仿真环境，定义状态空间、动作空间与奖励函数；继而设计 DQN 网络，以 ε-greedy 策略收集经验并更新 Q 表；最终训练收敛，模型在测试回合中稳定输出最优策略，性能曲线显著提升。

This personal small-scale project focuses on reinforcement learning. First, a simulation environment is constructed using Gym, with the state space, action space and reward function defined. Subsequently, a Deep Q-Network (DQN) is designed, which adopts the ε-greedy strategy to collect experience and update the Q-table. Finally, the training converges, and the model stably outputs the optimal policy during test episodes, with a notable improvement in the performance curve.

搜集汇总

背景与挑战

背景概述

该数据集基于强化学习项目，使用Gym仿真环境定义了状态空间、动作空间和奖励函数，并采用DQN网络与ε-greedy策略进行经验收集和Q表更新，最终训练收敛，模型在测试中实现稳定最优策略和性能提升。数据集包含一个333.00MB的MA-AT.zip文件，上传于2025年9月2日。