RoadLab: Intelligent advanced driving assistance system

1896年，第一起与驾驶有关的伤亡事件发生在英国伦敦。自那以后，世界卫生组织(WHO)估计，有2500万人死于与车辆有关的事故。世界范围内的伤害死亡人数预计将从1990年的510万人增加到2020年的840万人，其中与交通有关的伤害是造成这一增长的主要原因。这项研究的目的是通过首先加深对驾驶认知任务的理解，然后开发一个模型，将这些发现整合到预测模型中，从而减少这些死亡事件。这个模型的其他组成部分将是来自汽车本身的信息(车载仪器)和从汽车本身的传感器获得的关于环境的信息，如GPS、相机、雷达等)。这些不同的信息来源将被用来建立司机意向的预测模型。这项研究的主要目标是确定是否可以通过对驾驶员行为和来自汽车及其环境的信息的定量和定性分析来预测驾驶员的意图和与驾驶相关的行为。当然，这意味着为了得到有效的结果，这些分析可以在几分之一秒内完成。这需要开发用于处理数据和预测的智能算法，以及使用专门的高性能处理器；后者的原型已经开发出来。我们的主要研究目标是及时预测驾驶员的意图，以便警告驾驶员，或者潜在地通过自动驾驶干预来纠正驾驶员的行为。该计划的长期目标是解决导致开发智能、先进驾驶辅助系统的核心问题。

The first driving-related casualty was recorded in London, England in 1896. Since then, the World Health Organization (WHO) estimated that 25 million people died in vehicle-related accidents. World-wide deaths from injuries are projected to rise from 5.1 million in 1990 to 8.4 million in 2020, with traffic-related injuries representing the major cause for this increase. This research aims at reducing these fatalities by first developing a deeper understanding of the cognitive task of driving and then developing a model integrating these findings into a predictive model. The other components of this model will be information from the car itself (onboard instrumentation) and information about the environment obtained from sensors on the automobile itself, such as GPS, cameras, radar, etc.). These diverse sources of information will be used to build predictive models of driver intentionality. The primary goal of the research is to determine whether driver intentionality and driving-related actions can be predicted from quantitative and qualitative analyses of driver behavior and information from the car and its environment. This implies, of course, that these analyses can be done in a few fractions of a second in order to have effective results. This requires the development of clever algorithms for processing data and for prediction as well as the use of specialized high performance processors; a prototype of the latter has already been developed. Our main research goal is the timely prediction of driver intentionality in order to warn the driver or, potentially, to correct driver behavior through automated driving intervention. The long term goal of this program is to address core problems leading to the development of intelligent, advanced driving assistance systems.