Personalized Driver/Vehicle Lane Change Models for ADAS
Personalized Driver/Vehicle Lane Change Models for ADAS
复制标题
DOI：
10.1109/tvt.2014.2369522
复制
发表时间：
2015-10-01
影响因子：
6.8
通讯作者：
Ioannou, Petros
中科院分区：
文献类型：
--
作者：
Butakov, Vadim A.;Ioannou, Petros
车道变换对于驾驶员来说是紧张的操作，特别是在高速交通流期间。先进的驾驶辅助系统（ADAS）旨在帮助驾驶员在变道操纵。出于安全原因，为普通驾驶员或所有驾驶员开发的系统必须保守，以涵盖所有驾驶员/车辆类型。这种保守的系统对于激进的驾驶员可能是不可接受的，并且可能被更被动的驾驶员认为太激进。在变道操纵过程中考虑每个单独车辆/驾驶员系统的动力学和特性的ADAS将更有效，更容易被驾驶员接受，而不会牺牲安全性。在本文中，我们开发了一种方法，学习之前和车道变化过程中，在不同的驾驶环境下的个人驾驶员/车辆响应的特性。这些特征由一组模型捕获，这些模型的参数被在线调整以适应车道变化期间的个体车辆/驾驶员响应。我们开发了一个两层模型来描述机动运动学。下层描述车道变换作为一个运动学模型。高层模型建立了特定驾驶员的运动学模型参数值，并表示它们对周围车辆配置的依赖性。所提出的建模框架可以用作ADAS的核心组件，为驾驶员提供更个性化的建议，从而增加ADAS更广泛接受和使用的可能性。我们使用实际车辆和三个不同的驱动程序来评估所提出的方法。我们证明了该方法是有效的，在车道变换过程中，通过显示模型输出和原始数据之间的匹配的一致性，在建模个人驾驶员/车辆的响应。
Lane changes are stressful maneuvers for drivers, particularly during high-speed traffic flows. Advanced driver-assistance systems (ADASs) aim to assist drivers during lane change maneuvers. A system that is developed for an average driver or all drivers will have to be conservative for safety reasons to cover all driver/vehicle types. Such a conservative system may not be acceptable to aggressive drivers and could be perceived as too aggressive by the more passive drivers. An ADAS that takes into account the dynamics and characteristics of each individual vehicle/driver system during lane change maneuvers will be more effective and more acceptable to drivers without sacrificing safety. In this paper, we develop a methodology that learns the characteristics of an individual driver/vehicle response before and during lane changes and under different driving environments. These characteristics are captured by a set of models whose parameters are adjusted online to fit the individual vehicle/driver response during lane changes. We develop a two-layer model to describe the maneuver kinematics. The lower layer describes lane change as a kinematic model. The higher layer model establishes the kinematic model parameter values for the particular driver and represents their dependence on the configuration of the surrounding vehicles. The proposed modeling framework can be used as a kernel component of ADAS to provide more personalized recommendations to the driver, increasing the potential for more widespread acceptance and use of ADAS. We evaluated the proposed methodology using an actual vehicle and three different drivers. We demonstrated that the method is effective in modeling individual driver/vehicle responses during lane change by showing consistency of matching between the model outputs and raw data.