170 likes | 288 Vues
Algorithm for complete sensor coverage of complex 3D structures by autonomous agents, with focus on ship-hull inspection using bathymetric sonar. Utilizes closed triangular mesh model. Related work includes Art Gallery Problem, Car painting/Building inspection, and challenges in Computational Geometry. Presents a Sampling-based planning approach for complete coverage and feasible tour. Incorporates Postman Framework for deterministic point robots' inspections. Performance conclusion highlights efficiency and coverage over arbitrary 3D structures. Improvements sought in search for views and division between graph construction and path-finding.
E N D
Inspection Planning for Sensor Coverage of 3d Marine Structures Brendan Englot and Franz Hover Presented by Arvind Pereira for the CS599 Class on Sequential Decision Making in Robotics
Objective • Algorithm to achieve complete sensor coverage of complex 3-d structures when surveyed by an autonomous agent with multiple degrees of freedom • Motivating application is ship-hull inspection using bathymetric sonar • Plans on a closed triangular mesh model of the structure being inspected
Related Work • Art Gallery Problem – Minimum # of guards who can together observe the entire gallery • Car painting/Building inspection – Specific geometry of structure being covered • Challenging issues in Computational Geometry for complex 3-d models in the absence of a priori knowledge
Related Work • Sampling-based planning for similar problems employed the “art gallery” approach i.e. choosing a set of configurations which offer complete coverage and finding a feasible tour among these nodes • Postman Framework – information efficient approach to solve deterministically-designed inspections for point robots in a 2-d workspace when robot has a small FOV relative to the size of the workspace
Path-Finding Algorithm Incidence Matrix for flows Constraint Matrix ensuring that all nodes are inspected by a chosen path Constraint that enforces the initial node is on the graph
Conclusion • Algorithm ensures coverage planning over arbitrary discrete 3D structures using an information-efficient postman formulation • Need to be more efficient in the search for feasible views • Want to optimally manage the division between graph construction and path-finding