The construction industry is estimated to make up $400 billion in the US economy, employing over 5.5 million workers. However, construction work often suffers from low efficiency, where reworks due to defects, quality deviations or construction errors result in over almost $75 billion in wasted costs. Clearly, robotics and automation holds the key for the future of construction in order to deliver building projects in a way that is more accurate and efficient compared to conventional labor-intensive methods. This project proposes a framework for construction robots to scan and create schematic maps for highly dynamic and rapidly changing construction environments. Building accurate maps of construction sites is of great importance from both a robotics perspective and a construction management perspective. From a robotics perspective, semantic map building empowers robots with the capability to navigate and understand complex, dynamic workspaces. From a construction management perspective, having updated maps of the jobsite offers improved visibility in automating and scheduling construction projects and the capability to run construction work simulations and “what-if” scenarios. This project involves a confluence of research from robotics, machine learning, architecture, and civil and construction engineering. The overarching research goal is to achieve long-horizon map building, motion planning, and simulation in dynamic, unstructured construction sites through a merging