Robot Planning for Active Collision Avoidance in Modular Construction: Pipe Skids Example

Abstract

Robot-assisted assembly has shown great potential in modular construction for the future. However, as the complexity of prefabricated structural modules increases, the likelihood of unexpected robotic collisions also rises, such as unexpected collisions between robots and structural modules during the assembly process. Most robotic collision avoidance methods rely on known robot specifications, especially the work envelope, that is, the range profiles of movement created by a robot arm moving forward, backward, up, and down. In contrast, it is unknown what models of robots will be applied for future modular construction because of the rapid development of robot mechanical designs. The deep uncertainties related to future robot work envelopes can challenge any effort for planned robotic collision avoidance for robot-assisted modular construction. There is a pressing need for a simulation-based robot collision avoidance planning method for future robot-assisted modular construction that captures the deep uncertainties of work envelopes of future robots. As a result, this paper presents an active robotic collision avoidance method, called collision-free workspace and collision-avoidance path planning (CWCP), to tackle the unique robotic collision challenges in modular construction. First, CWCP uses an inverse kinematics Monte Carlo (IKMC) simulation in the design phase of a pipe skid module to scan a large number of possible robotic movement trajectories and the corresponding boundaries of the robotic space. This result can be used to identify a robust structural design with enough clearance for most robotic actions. Then, immediately before an operation, physics engine simulation is used to optimize the waypoints of the robotic arm movement to further eliminate the collision possibility given the established structural design and robotic specifications. The proposed method was tested with a simulated pipe skid installation task. The result showed that the proposed method helped strike a balance between a safe-tolerant design and robotic planning on site and thus is more suited for future robot applications with no prior knowledge about the work envelope specifications.