Human-Robot Integration for Pose Estimation and Semi-Autonomous Navigation on Unstructured Construction Sites

Compared to widespread successful deployment of robotic manipulators for repetitive and hazardous tasks in related industries such as manufacturing, the construction industry has achieved relatively limited benefits from robotics and soft automation. Unlike manufacturing, where robotic solutions benefit from the structured layout of the environment (e.g., factory assembly line), construction robots face unique challenges that arise from the rugged, dynamic, and unstructured environment of the work site, as well as the uncertainty and evolving sequence of occurring on-site events. This challenges any intended construction robots to not only replicate basic human motion, but also be capable of sensing and adapting to environmental changes, and making decisions based on the evolving state of the environment. Building upon recent advancements in robotic mapping, computer vision, and object recognition, the authors propose to introduce autonomous behavior at the basic task level for on-site construction robots to address these challenges in a flexible and extensible manner. This paper reports the outcome of the first phase of this research - a structured methodology for improved design and development of basic task automations - and focuses on algorithms developed for mobile robot navigation and relative pose estimation. The algorithms are implemented on a prototype mobile robotic platform, and evaluated in several experimental scenarios.