The increasing use of robotic systems in engineering offers great potential for automating tasks that are still largely performed manually, such as the quality inspection of buildings. Manual inspection is associated with safety risks, especially on active construction sites, where uneven terrain, cluttered environments, and dynamic obstacles such as workers and machinery are present. It is also time-consuming, difficult to standardise, and prone to human error. These limitations motivate the development of autonomous robotic solutions that can carry out inspection tasks safely, efficiently, and reliably. Quadruped robots are particularly well suited for this purpose due to their agility, manoeuvrability, and ability to operate on uneven and challenging terrain.

The aim of this PostDoc project was to establish the methodological foundations for autonomous navigation of a quadruped robot as a key prerequisite for robotic quality inspection in construction environments. Autonomous navigation consists of localisation, path planning, and control. The main focus of the project was the tight integration of localisation and path planning into a unified and mathematically consistent navigation framework that allows frequent and simultaneous updates. This is essential in dynamic and safety-critical environments, where delays between localisation and path planning can lead to hazardous situations, such as collisions. In addition, the framework was designed to provide stochastic information about the system state, enabling uncertainty-aware inspection results and the derivation of confidence measures.

This research was funded by the DFG until November 2025 as part of the i.c.sens Research Training Group (RTG 2159).