Search for: All records

Autonomous navigation in construction environments is particularly challenging due to dynamic obstacles and uncertain surroundings. While recent advances in Building Information Modeling (BIM)-based planning have leveraged spatial and semantic information to improve navigation, most prior work assumes precise localization of the BIM model to enable global path planning. In contrast, this paper introduces an online replanning framework that registers obstacles on discovery within BIM and replans according to the updated semantic map. Our method integrates object-aware path planning by utilizing large language models (LLMs) to extract semantic danger sentiments from BIM-annotated objects and their spatial information about the mission environment. Additionally, we demonstrate practical feasibility by integrating a path tracking control, ensuring generated paths are not only safer but also realistically executable by mobile robots. Experimental results demonstrate an improved obstacle avoidance by 2.8× compared to traditional A* algorithms in dynamically updated environments. 

Quadrupeds are becoming increasingly popular in construction engineering research and practice for their affordability and accessibility. These robots navigate uneven terrain commonly found in construction sites, making them suitable vehicles for sensors and monitoring tasks. However, the lack of streamlined and fully developed client-side software packages inhibits rapid deployment of application-specific models to the field. Furthermore, substantial prerequisite knowledge of computer science and programming significantly impedes the ability of non-experts to adapt the robots to specific applications. In this work, we present a comprehensive framework to address these gaps in accessibility, enabling users to customize these robots to their needs. This framework provides a template that facilitates seamless communication between the robotic vehicle, edge devices, sensors, pathfinding algorithms, and a Unity simulation for mission planning and execution. As an example of this framework’s flexibility, we have conducted a case study using this template to demonstrate an application of the framework in the construction domain that performs worker activity recognition and features a novel self-labeling mechanism for construction activity video data. The findings highlight the potential of accessible software tools in expanding the utility of robotic platforms across various engineering domains. 

Robots have the potential to enhance safety on construction job sites by assuming hazardous tasks. While existing safety research on physical human-robot interaction (pHRI) primarily addresses collision risks, ensuring inherently safe collaborative workflows is equally important. For example, ergonomic optimization in co-manipulation is an important safety consideration in pHRI. While frameworks such as Rapid Entire Body Assessment (REBA) have been an industry standard for these interventions, their lack of a rigorous mathematical structure poses challenges for using them with optimization algorithms. Previous works have tackled this gap by developing approximations or statistical approaches that are error-prone or data-dependent. This paper presents a framework using Reinforcement Learning for precise ergonomic optimization that generalizes to different types of tasks. To ensure practicality and safe experimentations, the training leverages Inverse Kinematics in virtual reality to simulate human movement mechanics. Results of a comparison between the developed framework and ergonomically naive approaches are presented.