Studies in machine learning and robotics for automated construction

这项博士研究的目的是通过利用传感器数据融合和机器学习(ML)方面的最新进展，实现地下数字双胞胎的构建方式的阶段性变化。将使用数字地面测量技术(例如，探地雷达、地震扫描仪、照相机和激光雷达)在施工期间开发地下空间的3D地图。将开发一个可扩展的ML框架，以实现准确的3D重建并执行感知、预测和规划任务。融合后的感官输入随后将被用来构建数字孪生兄弟。开发的框架将通过行业合作伙伴和Infratech初创企业HyperTunes Ltd.提供的真实数据进行开发和验证。

The aim of this PhD research is to achieve a step-change in how underground digital twins are constructed by leveraging recent advances in sensor data fusion and machine learning (ML). Digital ground surveying techniques (e.g. ground penetrating radars, seismic scanner, camera and Lidar) will be used to develop 3D mapping of the underground space during construction. A scalable ML framework will be developed to achieve accurate 3D reconstruction and perform perception, prediction and planning tasks. The fused sensory inputs will subsequently be used to build the digital twin. The developed framework will be developed and validated through real-world data provided by industry partner and infraTech start-up hyperTunnel Ltd.