With the emergence of vehicle-communication technologies, many researchers have strongly focused their interest in vehicle energy-efficiency control using this connectivity. For instance, the exploitation of preview traffic enables the vehicle to plan its speed and position trajectories given a prediction horizon so that energy consumption is minimized. To handle the strong uncertainties in the traffic model in the future, a constrained controller is generally employed in the existing researches. However, its expensive computational feature largely prevents its commercialization. This paper addresses computational burden of the constrained controller by proposing a computationally tractable model prediction control (MPC) for real-time implementation in autonomous electric vehicles. We present several remedies to achieve a computationally manageable constrained control, and analyze its real-time computation feasibility and effectiveness in various driving conditions. In particular, both warmstarting and move-blocking methods could relax the computations significantly. Through the validations, we confirm the effectiveness of the proposed approach while maintaining good performance compared to other alternative schemes.

中文翻译：

---

基于计算廉价模型预测控制的无人驾驶电动汽车电池能量管理

随着车辆通信技术的出现，许多研究人员将注意力集中在使用此连接的车辆能效控制上。例如，预览交通的利用使车辆能够在给定预测范围的情况下计划其速度和位置轨迹，从而将能耗降至最低。为了应对未来交通模型中的强烈不确定性，现有研究中通常采用约束控制器。但是，其昂贵的计算功能在很大程度上阻碍了其商业化。本文提出了一种可计算的易于处理的模型预测控制（MPC），用于自动电动汽车中的实时实现，从而解决了约束控制器的计算负担。我们提出了几种方法来实现可计算管理的约束控制，并分析其在各种驾驶条件下的实时计算可行性和有效性。特别是，热启动和移动阻止方法都可以显着放松计算。通过验证，与其他替代方案相比，我们确认了所提出方法的有效性，同时保持了良好的性能。