A Switched Systems Approach for Navigation and Control withIntermittent Feedback
Abstract
Imaging systems provide unique sensing capabilities for estimation and control. Image feedback typically involvestracking the evolution of image features in time/space to determine the relative motion of an autonomous systemor tracked targets. The resulting image dynamics are nonlinear and include uncertainties inherent to the Euclideanspaceto image-space mapping. Moreover, such feedback is intermittent due to the inevitable loss of features due to thelimited camera field-of-view or feature occlusion. The intermittent nature of image feedback can lead to degradation orinstabilities in developed state estimates or controllers. Efforts in this project seek to generalize the typical assumptionof continuous feature point observation from a single imaging source for estimation and control problems. To developestimation and control solutions that use image-based feedback, fundamental problems associated with analyzing thestability of uncertain switched nonlinear systems must be addressed. The fundamental contribution of Aim 1 includesthe development of switched systems tools using Lyapunov-based to enable arbitrary switching between differentimage sources (i.e., synthetic persistence where switching is between stable subsystems) using asymptotic adaptiveupdate laws. In Aim 2, a novel data-based integral concurrent learning method is proposed to yield exponentiallearning to facilitate the development of dwell-time conditions for switching between stable (when features are visible)and unstable (when persistence is lost) subsystems. Efforts in Aim 3 focus on the novel idea of developing dwell-timedependent controllers/path planners that can purposefully exploit the flexibility permitted by not having to maintaincontinuous views of a target/reference frame. An anticipated outcome of this project is new design/analysis tools thatinvolve adaptation for switched nonlinear systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 09, 2018
- Source ID
- FA95501810109
Entities
People
- Warren E Dixon
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Florida