Spatio-Temporally Smooth Local Mapping and State Estimation Inside Generalized Cylinders with Micro Aerial Vehicles

In this letter, we consider state estimation and local mapping with a micro aerial vehicle inside a tunnel that can be modeled as a generalized cylinder, using a three-dimensional lidar and an inertial measurement unit. This axisymmetric environment poses unique challenges in terms of localization and mapping. The point cloud data returned by the sensor consists of indiscriminate partial cylindrical patches complicating data association. The proposed method reconstructs an egocentric local map through an optimization process on a nonlinear manifold, which is then fed into a constrained unscented Kalman filter. The proposed method easily adapts to different diameters, cross sections, and changes in center line curves. The proposed approach outperforms our previous contribution [T. Ozaslan, G. Loianno, J. Keller, C. J. Taylor, V. Kumar, J. M. Wozencraft, and T. Hood, 'Autonomous navigation and mapping for inspection of penstocks and tunnels with MAVs,' IEEE Robotics Automation Letter, vol. 2, no. 3, pp. 1740-1747, Jul. 2017] in terms of mapping quality and robustness to noncylindrical cross sections. Our motivation is to automate the labor intensive, dangerous, and the expensive inspection of penstocks with the least operator intervention. We present experimental results obtained in Center Hill Dam, TN, USA, to validate the proposed approach.