Cooperative Terrain-relative Navigation

Download or read book Cooperative Terrain-relative Navigation written by Adam Tadeusz Wiktor and published by . This book was released on 2021 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: This thesis introduces a new method to improve localization performance for teams of vehicles navigating cooperatively. When fusing measurements between multiple vehicles, the structure of the cooperative navigation network inherently introduces correlation between them, causing many traditional filter architectures to overconverge and become inconsistent. The algorithm presented in this thesis addresses this correlation and properly fuses measurements, allowing improved performance over other existing methods while still guaranteeing consistency. When restricted to linear, Gaussian systems, the covariance recovers 99% of the performance of an ideal centralized filter in some tests. Additionally, a proof is presented to guarantee that the algorithm is consistent under standard Kalman filter assumptions. The algorithm is also extended to apply to nonlinear systems, losing the guarantees of consistency (as with all Kalman filters) but achieving good performance in practice. This allowed the method to be tested in a laboratory experiment with real-world sensors. Finally, this thesis further extends the algorithm to apply to non-parametric particle filters, allowing for full cooperative Terrain-Relative Navigation (TRN) with multi-modal position estimates. This is demonstrated in simulation, where cooperative TRN is shown to provide a 63% reduction in localization error over standard single-vehicle TRN for one example mission, reducing the average error from 23.7m to 8.7m for a vehicle over flat terrain. The cooperative TRN algorithm is also demonstrated using field data from a team of Long-Range Autonomous Underwater Vehicles in Monterey Bay. In offline testing, the cooperative TRN method was able to correctly find the position of a vehicle when its own individual TRN filter was unable to converge. This demonstrates that the cooperative TRN algorithm is effective with real-world robotic systems, increasing localization accuracy and enabling new missions involving navigation in flat, unmapped, or changed terrain.