以期刊论文的格式将以下句子修改通顺：在测试阶段将DQN、DDQN和Dueling DDQN的权重参数固定为训练10000回合时的权重参数并将固定权重参数的三种模型集成在输出端增加集成层构成集成深度强化学习模型。记录上述三种网络模型和集成深度强化学习模型在仿真环境中运行100回合的车辆驾驶行为决策数据如图10和表4所示图10为不同模型在100个测试回合内的驾驶行为统计表4为不同模型的驾驶行为决策成功

During the testing phase, the weight parameters of DQN, DDQN, and Dueling DDQN were fixed to those obtained after 10000 training rounds. The three models with fixed weight parameters were integrated by adding an ensemble layer at the output, resulting in an integrated deep reinforcement learning model. We recorded the vehicle driving decision data for each of the three network models and the integrated deep reinforcement learning model, which were tested in a simulation environment for 100 rounds. Figure 10 and Table 4 show the statistical results of driving behavior for different models over 100 testing rounds. Figure 10 depicts the driving behavior of different models over 100 testing rounds, while Table 4 shows the success rate of driving behavior decision-making, average vehicle speed, and average consumption time for each model. The success rate of driving behavior decision-making was calculated as the ratio of successful testing rounds to the total testing rounds, with a successful round being defined as a round that terminated under condition (3) of the simulation.