One of the main challenges for road authorities is to maintain the quality of the road infrastructure. Road anomalies can have a significant impact on traffic flow, the condition of vehicles, and the comfort of occupants of vehicles. Strategies such as pavement management systems use pavement evaluation vehicles that are equipped with state-of-the-art devices to assist road authorities in identifying and repairing these anomalies. The quantity of data available is limited, however, by the limited availability and, therefore, coverage of these vehicles. To address this problem, several investigations have been conducted on the use of smartphones or equipping vehicles with additional sensors to identify the presence of road anomalies. This paper aims to add to this arsenal by using sensors already available in production vehicles to identify road anomalies. If production vehicles could be used to identify road anomalies, then road authorities would be equipped with an additional fleet of mobile sensors (vehicles traveling on a particular road) to receive initial insights into the presence of anomalies. This information could then be used to assist road authorities to deploy their staff and equipment more precisely at these locations, such that appropriate equipment reaches the right place at the right time. In this paper, an algorithm that uses lateral acceleration and individual wheel speed signals, which are commonly available vehicular variables, was developed to detect potholes using machine learning techniques. The results of the algorithm were validated with real life test scenarios.

道路当局面临的主要挑战之一是保持道路基础设施的质量。道路异常会严重影响交通流量，车辆状况以及车辆乘员的舒适度。诸如路面管理系统之类的策略使用配备有最新设备的路面评估工具，以协助道路管理部门识别和修复这些异常情况。然而，可用数据的数量受到这些车辆的有限可用性和覆盖范围的限制。为了解决这个问题，已经进行了一些关于使用智能手机或为车辆配备附加传感器以识别道路异常情况的调查。本文旨在通过使用量产车辆中已经可用的传感器来识别道路异常，从而增加这个武器库。如果可以使用量产车来识别道路异常，则道路当局将配备额外的移动传感器车队（在特定道路上行驶的车辆），以初步了解异常情况。然后，可以使用此信息来协助道路管理部门在这些位置更准确地部署其人员和设备，以使合适的设备在正确的时间到达正确的位置。在本文中，使用机器学习技术开发了一种算法，该算法使用横向加速度和单个车轮速度信号（通常是车辆变量）来检测坑洼。