A new class of estimation methods are developed that can optimally utilize minima resources for high-precision pose tracking. This project develops the first formal methodologies for providing selections of parameters (e.g., camera frame rate, image resolution, number and type of detected features) for small portable devices to delay depletion of the battery. The approach is based on a rigorous study of the properties of the pose tracking problem with visual and inertial sensors and (1) identifies the fundamental limits of the attainable estimation accuracy, and (2) allows the analytical prediction of the accuracy as a function of the use of the sensing and processing resources. This makes possible the development of algorithms that optimally allocate system resources whose design relies on an optimization framework where the estimation errors constitute the cost function to be minimized, the resource limitations are explicitly modeled as constraints, and all the relevant design parameters (e.g., camera frame rate, number of features used) are the optimization variables. The immediate impact of this research effort is the increased cost efficiency and accuracy for navigation tasks in diverse applications.
The developed technology is available to the wider community through open-source position-tracking software for mobile phone devices and provides useful assistive technology. We engage K-12 students at local outreach events and introduce them to engineering as well collaborate with MESA, a long-standing program at the University of California at Riverside with a proven record of attracting underrepresented minority students to science and engineering.