F5-T DeepSLAM: Visual Intelligence for Navigation and Planning

PIs: Warren Dixon, Andy Li

The goals of this project include developing the ability to use offline, transferred, and real-time learning using various data sources (intermittent state feedback, cloud) to enable SLAM and related image-based estimation methods. This goal includes developing a means to integrate learned information at different timescales and data sources to increase the robustness of the composite system beyond the individual capabilities of each subsystem to facilitate navigation, path-planning, and control of autonomous vehicles. Specific objectives include: 1) develop deep learning methods (DeepSLAM) to enable enhanced image-based feature tracking and automatic target recognition, 2) develop real-time image-based target estimators and path planning and control methods that allow the target/features to intermittently leave the field of view, 3) develop associated dwell-time condition that can quantify the reliability of the developed estimators, 4) incorporate data-based learning (concurrent learning and Deep learning) to extend the dwell-time (provide greater robustness), and 5) implement the developed methods on an autonomous air or ground vehicle to demonstrate the performance of the method.