求翻译：建立了定周期和变周期两种模式下的离线Q学习模型，分别针对等饱和度和延误最小的两种优化目标进行研究，建立了等饱和度目标的奖励函数。算例结果表明定周期等饱和度Q学习模型的最优解唯一，定周期延误最小Q学习模型的部分解唯一，变周期Q学习模型的解均不唯一。对于最优解不唯一的解结构，可以将这些解作为一个最优解的方案库，当检测器检测到交通流量时，从方案库中进行选择。这时可以考虑与上一个配时方案周期接近、与相邻交叉口周期接近等因素，提高配时方案与其他因素的兼容性、鲁棒性，这是传统配时理论不具备的。是什么意思？

正在翻译，请等待...

There will be a constant cycle and cycle two modes, the off-line Q learning models, respectively, at saturation and delays, the smallest of the two optimization goals for research, and the establishment of the saturation of the target award functions. Results indicate that there will be examples of

Established a cycle and cycle two of offline mode Q-learning models, respectively for saturation and delay a minimum of research on two kinds of optimization goal, the establishment of a saturation of targets such as award functions. Example shows cycle saturation, such as Q-learning model of optima