A project report on coverage path planning for structural inspection, focused on exploring an unknown model rather than the region around it.
Abstract
Using UAVs for structural and model inspection is a well-established area of research, but most existing work focuses on completely exploring an unknown region, the space the UAV flies through. This project instead focuses on exploring the unknown model itself: the actual structure being inspected, which is often the thing that matters more for inspection use cases.
System components
- A virtual sensor simulating real-world sensing constraints
- A UE4 (Unreal Engine 4) simulation environment
- A global path planner for overall coverage strategy
- A local path planner using a novel coverage-ratio approach to decide, moment to moment, whether continuing to explore the model is worth more than exploring the surrounding region
Results
Simulated in AirSim, the approach covered roughly 70% of the target model, generating an average of 250 poses per run, a meaningful result given the deliberate focus on model coverage over blanket regional exploration.
Future directions
Better pose-selection criteria (considering multiple candidate poses instead of committing to one), recalculating the global path as the unknown region is traversed, and faster computation to reduce total mission time.
Mohammad Hameed, University of Calgary. CPSC 789 Project Final Report.