Department of Automotive and Mechatronics Engineering, Ontario Tech University, Oshawa, ON L1G 0C5, Canada.
Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada.
Sensors (Basel). 2021 Jan 7;21(2):364. doi: 10.3390/s21020364.
Excavation is one of the primary projects in the construction industry. Introducing various technologies for full automation of the excavation can be a solution to improve sensing and productivity that are the ongoing issues in this area. This paper covers three aspects of effective excavation progress monitoring that include excavation volume estimation, occlusion area detection, and 5D mapping. The excavation volume estimation component enables estimating the bucket volume and ground excavation volume. To achieve mapping of the hidden or occluded ground areas, integration of proprioceptive and exteroceptive sensing data was adopted. Finally, we proposed the idea of 5D mapping that provides the info of the excavated ground in terms of geometric space and material type/properties using a 3D ground map with LiDAR intensity and a ground resistive index. Through experimental validations with a mini excavator, the accuracy of the two different volume estimation methods was compared. Finally, a reconstructed map for occlusion areas and a 5D map were created using the bucket tip's trajectory and multiple sensory data with convolutional neural network techniques, respectively. The created 5D map would allow for the provision of extended ground information beyond a normal 3D ground map, which is indispensable to progress monitoring and control of autonomous excavation.
挖掘是建筑行业的主要项目之一。引入各种技术实现挖掘的完全自动化，可以解决该领域中持续存在的传感和生产力问题。本文涵盖了有效挖掘进度监测的三个方面，包括挖掘体积估计、遮挡区域检测和 5D 映射。挖掘体积估计组件可用于估算铲斗体积和地面挖掘体积。为了实现对隐藏或遮挡地面区域的映射，采用了本体感受和外部感受传感数据的集成。最后，我们提出了 5D 映射的概念，该概念使用带有激光雷达强度和地面电阻指数的 3D 地面地图，以几何空间和材料类型/特性的形式提供已挖掘地面的信息。通过使用迷你挖掘机进行实验验证，比较了两种不同的体积估计方法的准确性。最后，使用铲斗尖端的轨迹和多个具有卷积神经网络技术的传感器数据分别创建了遮挡区域的重建地图和 5D 地图。创建的 5D 地图可以提供超出普通 3D 地面地图的扩展地面信息，这对于自主挖掘的进度监测和控制是必不可少的。
