Show simple item record

dc.contributor.authorSchoenberg, Jonathanen_US
dc.date.accessioned2012-06-28T20:57:12Z
dc.date.available2017-06-01T06:00:27Z
dc.date.issued2012-01-31en_US
dc.identifier.otherbibid: 7745235
dc.identifier.urihttps://hdl.handle.net/1813/29366
dc.description.abstractThis thesis explores data fusion and distributed robotic perception through a series of theoretical developments, analyses and experiments. First, a GSF with component extended Kalman filters (EKF) is proposed as an approach to localize an autonomous vehicle in an urban environment with limited GPS availability. The GSF is used because of its ability to represent the posterior distribution of the vehicle pose with better efficiency (fewer terms, less computational complexity) than a corresponding bootstrap particle filter with various numbers of particles due to the interaction with measurement hypothesis tests. A series of in-depth empirical studies are performed using 37 minutes of recorded data from Cornell University's autonomous vehicle driven in an urban environment, including a 32 minute GPS blackout. Second, a distributed grid-based terrain mapping algorithm using Gaussian Mixture Models is developed for use in tree connected and arbitrary connected sensor networks. The distributed data fusion rules are developed that operates directly on the sufficient statistics summarizing the grid-cell height and uncertainty. The distributed grid-based terrain mapping algorithms is demonstrated in an experimental environment involving 8 autonomous robots operating in an indoor environment for 120 seconds. Third, an algorithm to segment 3D points in dense range maps generated from the fusion of a single optical camera and a multiple emitter/detector laser range finder is presented. The algorithm is demonstrated on data collected with the Cornell University DARPA Urban Challenge vehicle. Finally, two information theoretic procedures for fusing multiple distributions with unknown correlation are developed. The first approach developed is Entropy Weighted Chernoff fusion; this fusion procedure biases the WEP fusion weight towards the distribution with the lowest entropy. An information loss for the WEP conservative fusion rule is introduced and an approximation derived by computing the Kullback-Leibler divergence between the Naive Bayes and WEP fused distributions. The approximation is minimized for the second fusion approach: Minimum-Information-Loss fusion; the procedure generates the least conservative fused distribution in the family of WEP results. Experimental results include the fusion of multiple occupancy grid maps over an optimally connected sensor network, demonstrating consistent map estimates.en_US
dc.language.isoen_USen_US
dc.subjectDistributed Data Fusionen_US
dc.subjectRobotic Perceptionen_US
dc.subjectEstimationen_US
dc.titleData Fusion And Distributed Robotic Perceptionen_US
dc.typedissertation or thesisen_US
thesis.degree.disciplineAerospace Engineering
thesis.degree.grantorCornell Universityen_US
thesis.degree.levelDoctor of Philosophy
thesis.degree.namePh. D., Aerospace Engineering
dc.contributor.chairCampbell, Marken_US
dc.contributor.committeeMemberPsiaki, Mark Lockwooden_US
dc.contributor.committeeMemberKoutsourelakis, Phaedon-Steliosen_US
dc.contributor.committeeMemberKress Gazit, Hadasen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record

Statistics