Introduction

Regular inspection and monitoring of aging assets are crucial to safe operation in industrial facilities, with remote robotic monitoring being a particularly promising approach for offshore asset inspection. However, vessels, pipework, and surfaces to be monitored can follow complex 3D surfaces, and frequently no 3D as-built models exist. In this paper, we present an end-to-end solution, that uses an optimisation method for coverage path planning of multiple complex surfaces, for mobile robot manipulators. The system includes a two-layer hierarchical structure of optimisation: mission planning and motion planning. The surface sequence is optimised with a mixed-integer non-linear programming formulation while motion planning solves a whole-body optimal control problem considering the robot as a planarly floating-base system. The loco-manipulation system automatically plans a full-coverage trajectory over multiple surfaces for contact-based non-destructive monitoring after unrolling the 3D-mesh region-of-interest selected from the user interface and projects it back to the surface. Our pipeline aims at offshore asset inspection and remote monitoring in industrial applications. We demonstrate the generality and scalability of our solution in a variety of robotic coverage path planning applications, including for multi-surface asset inspection using a quadrupedal manipulator. 

Robot model

Perceptive loco-manipulation system

The robot obtains environmental information with the onboard Realsense camera attached to the wrist of the arm, enabling autonomy in task and motion planning.

• Environment with a single, curved surface:

• Environment with two flat surfaces:

Multi-surface planning

Full pipeline demo