众包式路内泊位的移动探测研究

As the number of vehicles continues to grow, parking spaces are at a premium in city streets. Additionally, due to the lack of knowledge about on-street parking spaces, heuristic circling the blocks not only costs drivers' time and fuel, but also increases city congestion..After reviewing the high-profile smart parking systems at home and abroad, we found the sources why the states of on-street parking spaces are difficult to be obtained. Although each of the reviewed systems comes with different sensor technologies, e.g., sonar, infrared or magnetometer, all of them can be classified as the stationary sensing solutions, namely, burying sensors under or mounting sensors over the parking spaces. The strength of stationary sensing is the high detection accuracy, while the weak point is the heavy overall costs, which are 1) the ratio between the required number of sensors and the number of parking spaces is greater than or equal to 1; 2) it is not only expensive, but also unscalable to dig holes or mount overhead sensors for every on-street parking space; 3) comparing to off-street parking, the on-street one has more harsh environment, which may negatively impact sensors’ functionality. For these reasons, the success of those pilot smart parking systems can not be extended to the city wide..We propose a crowdsourcing mobile detection alternative: sonars, placed on the side doors of buses or taxis, continuously measure the distance from the vehicles to roadside while the vehicles are cruising on the streets. The sonar data, combined with GPS, speeds, weather, holidays and other factors, are analyzed by a supervised learning algorithm that can detect on-street parking occupancy and spot illegally-parked vehicles. The ratio between the required number of sensors and the number of parking spaces is much smaller than 1, while the proposed solution can still achieve the high accuracy compared with the fixed sensing one.

汽车数量的迅猛增长，让城市的路内泊位成了一种具有溢价效应的紧俏资源。大部分城市目前无法为车主提供路内泊位的状态信息，使车主只能在目的地附近探索性的绕圈寻找。.在对比国内外知名智慧城市停车系统后，我们找到了路内泊位状态难以获取的原因。虽然他们尝试了不同传感器技术（如声纳或磁强计），但都囿于定点探测的方法，即，在泊位的上或下方安装传感器。其优点是探测准确率高，缺点是综合成本高：1) 所需传感器数量与泊位比大于等于1；2) 传感器及其配套设施需要路面施工；3) 路面数据噪音多。这些原因导致定点探测式的路内泊位系统难以推广。.我们提出一种众包的移动探测方案：将声纳置于公交或出租车侧门，其在行驶过程中，不断测量车辆到路边的距离，结合GPS，车速，天气及节假日等因素，使用监督机器学习算法，找出路边空位或车辆违停的特征向量，训练出预测模型。在获得与定点探测系统相似的成功率情况下，传感器与泊位比远小于1。

本项目围绕路内众包泊位的移动探测解决方案的探索和优化。在对比国内外知名智慧城市停车系统后，我们找到了路内泊位状态难以获取的原因。我们提出一种众包的移动探测方案：将声纳置于公交或出租车侧门，其在行驶过程中，不断测量车辆到路边的距离，结合GPS，车速，天气及节假日等因素，使用监督机器学习算法，找出路边空位或车辆违停的特征向量，训练出预测模型。在获得与定点探测系统相似的成功率情况下，众包移动探测方案所需传感器与泊位比远小于1。.在项目实施期间，我们主要围绕国内外路边智能停车项目的调查与基于超声波传感器的移动众包路边泊位探测系统的建立两个大的方向开展。通过项目的执行，我们建立了移动众包路边泊位路试平台，获得了宝贵的众包路试数据；我们提出了理论研究模型，验证路试平台的数据，收获了6篇已发表文章；提交了2项发明专利申请，目前处于专利实质审查阶段；项目吸引并培养了多名学生参与，其中一名硕士和四名本科生实质参与项目研究，并发表论文及专利（含在审）。