横向互联空气悬架多智能体减振器系统博弈控制
Game control of multi-agent damper system for laterally interconnected air suspension

为进一步改善横向互联空气悬架车辆的行驶平顺性和操纵稳定性,基于多智能体理论和合作博弈Shapley值原理构建多智能体减振器控制系统; 多智能体减振器控制系统由信息发布智能体、平顺性智能体、操稳性智能体和博弈协调智能体组成,其中信息发布智能体从环境中获取车辆状态信息,根据下层智能体的信息需求传递信息,平顺性智能体接收悬架动行程及其变化率信息,根据平顺性控制要求,输出自身的阻尼系数意图,操稳性智能体接收当前互联状态信息触发对应的推理模块,根据车身侧倾角信息求解需求的阻尼系数,其中推理模块是通过对遗传算法优化出的阻尼系数进行模糊神经网络自学习形成的,博弈协调智能体接收平顺性智能体与操稳性智能体的阻尼意图,根据自身的合作博弈规则,对阻尼意图进行修正,输出全局最优阻尼系数; 在不同互联状态、不同激励条件下进行空气悬架静、动态特性试验研究,并将试验结果与仿真结果进行对比,验证仿真模型的准确性; 在混合工况下,利用整车仿真模型验证多智能体减振器控制系统的可行性和有效性。研究结果表明:和传统减振器阻尼控制系统相比,多智能体减振器控制系统能有效地使簧载质量加速度均方根值降低14.95%,悬架动行程均方根值降低10.64%,车身侧倾角均方根值降低12.33%。提出的多智能体减振器控制系统改善了车辆行驶平顺性和乘坐舒适性,并且能够抑制车身的侧倾,提高整车的操纵稳定性。
To further improve the ride comfort and handling stability of vehicles equipped with laterally interconnected air suspension(LIAS), on the basis of multi-agent theory and the cooperative game Shapley value principle, a multi-agent damper control system was constructed. The multi-agent vibration absorber control system was composed of an information publishing agent, ride comfort agent, handling stability agent, and game cooperation agent. The vehicle state information from the environment was obtained by the information publishing agent, and the information transmission was finished according to the information demand of the lower agent. The suspension dynamic travel and its changing rate information were received by the ride comfort agent, and its own damping coefficient intention was output according to the ride control requirements. The information of the current interconnected state was received by the handling stability agent, the corresponding reasoning module was triggered, and the required damping coefficient was solved according to the information of the carbody roll angle. The reasoning module was formed by fuzzy neural network self-learning to the damping coefficient optimized by a genetic algorithm. The damping intents of the ride comfort agent and handling stability agent were received by the game cooperation agent, the damping intents were modified according to its own cooperative game rules, and the global optimal damping coefficient was outputted. Under different interconnected states and different excitation conditions, the static and dynamic characteristics of the air suspension were tested and compared with the simulation results, and the accuracy of the simulation model was verified. Under the condition of mixed construction, the feasibility and effectiveness of the multi-agent damper control system were verified by a vehicle simulation model. Analysis result shows that compared with the traditional damping control system, the multi-agent damper control