Abstract-We present a hybrid algorithm that plans feasible paths for 100% sensor coverage of complex 3D structures. The structures to be inspected are segmented to isolate planar areas, and back-and-forth sweep paths are generated to view as much of these planar areas as possible while avoiding collision. A randomized planning procedure fills in the remaining gaps in coverage. The problem of selecting an order to traverse the elements of the inspection is solved by reduction to the traveling salesman problem. We present results of the planning algorithm for an autonomous underwater vehicle inspecting the in-water portion of a ship hull. The randomized configurations succeed in observing confined and occluded areas, while the 2D sweep paths succeed in covering the open areas.
I. INTRODUCTIONCoverage path planning enables fast and efficient task completion in applications that require an autonomous agent to sweep an end effector over some portion of its workspace, including sensing, cleaning, painting, and plowing [7]. Optimal coverage paths often utilize a back-and-forth sweeping motion to cover the required areas efficiently. This is achieved in obstacle-filled 2D workspaces using cell decomposition methods [6], [14], which allow areas of open floorspace to be swept with uninterrupted motions. In 3D workspaces, the coverage task typically requires a full sweep of the interior or exterior boundary of a 3D structure embedded in the workspace. Back-and-forth sweeping has achieved uniform coverage of curved surface patches [3], and circumferential looping around 2D cross-sections has been used to cover the full boundary of closed 3D structures [2], [5].The paths planned by these algorithms contain uniform spacing between tracklines and often accumulate data sliceby-slice along a single spatial dimension of the workspace. Travel along a highly regular inspection route of this type allows a human operator to monitor task completion, and facilitates easy reading and interpretation of a sensor-based data product. To our knowledge, however, the existence of an arbitratry, collision-free coverage path is not a sufficient condition for the existence of a route with uniform spacing or a layout along a single spatial dimension.On the other hand, covering the boundary of a 3D structure using randomly sampled view configurations, a technique that employs a discrete set of stationary views rather than a continuous sensing trajectory, [9], [10], has been shown