Multi-task Model Predictive Control based on Continuation with Intermediate Mode
Multi-task Model Predictive Control based on Continuation with Intermediate Mode
复制标题
基于中间模式延续的多任务模型预测控制
DOI：
10.1109/itsc45102.2020.9294663
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Suzuki Tatsuya
中科院分区：
文献类型：
--
作者：
Honda Kohei;Okuda Hiroyuki;Suzuki Tatsuya
针对自动驾驶（AD）中的特定任务，已经有许多使用模型预测控制（MPC）的同时路径规划和运动控制（SPMC）的研究。为了将MPC应用于“时变”的真实的环境，有必要在一致的控制架构下连续切换这些面向任务的MPC。MPC切换的关键问题之一是如何接管两个不同优化问题的可行性，实现任务的平滑切换。本文提出了一种新的切换MPC系统平滑过渡框架，可以减少不可行性。我们提出的框架利用了一个中间模式和过渡规则，以避免这种不可行性。它们可以通过简单地将每个面向任务的MPC与中间模式嵌入在一起来实现驾驶任务的平滑过渡。所提出的架构的有效性证明了数值模拟，解决了自适应巡航控制/车道保持（ACC/LK）和换道（LC）任务之间的任务转换。
There have been many researches on simultaneous path planning and motion control (SPPMC) for specific tasks in autonomous driving (AD) using model predictive control (MPC). In order to apply the MPC to `time-varying' a real environment, it is necessary to continuously switch these task-oriented MPCs under consistent control architecture. One of the critical issues in the switching of MPCs is how to take over the feasibility of the two different optimization problems so as to realize the smooth task switching. This paper presents a novel framework for smooth transitions in the switched MPC system which can reduce the infeasibility. Our proposed framework exploits an intermediate mode and the transition rule to avoid such infeasibility. They make it possible to get a smooth transition of driving tasks by simply embedding each task-oriented MPC together with the intermediate mode. The effectiveness of the proposed architecture is demonstrated by a numerical simulation that addresses the task transition between adaptive-cruise-control/lane-keep (ACC/LK) and lane-change (LC) tasks.