# Browsing by Subject "Gaussian mixture model"

Now showing 1 - 3 of 3

- Results Per Page
1 5 10 20 40 60 80 100

- Sort Options
Ascending Descending

Item Human detection, gesture recognition, and policy generation for human-aware robots(2017-05) Jorgensen, Steven Jens M.; Sentis, LuisShow more For robots to be deployable in human occupied environments, the robots must have human-awareness and generate human-aware behaviors and policies. This thesis posits that a human-aware robot must be capable of (1) human detection and tracking, (2) human action or intent recognition and (3) intelligent, human-aware action generation. This work presents and evaluates a methodology for each stated capability. In Chapter 2, a method for practical side-by-side human detection for the Valkyrie robot using the Multisense SL sensor is presented. An explanation of why current off-the-shelf techniques are not suitable and a depth-based algorithm using point cloud descriptors and a Random Forest classifier for detecting humans under occlusion, in close proximity, in varying sparsity, and in random poses on the Multisense SL sensor are presented. In Chapter 3, action recognition of arm motion gestures is framed as a supervised learning problem. A popular technique for gesture representation with dynamic movement primitives (DMPs) and its classification using Gaussian Mixture Models (GMMs) is explored. The approach is tested under various hypotheses to understand the intricacies of using DMPs for movement representation. The following findings are reported: (a) recognition rate is sensitive to the number of basis weights, (b) DMPs can be used to recognize two linear motions, (c) rhythmic gestures can be differentiated with the discrete formulation of DMPs, and (d) DMPs can represent static-type gestures. In Chapter 4, a novel technique for (a) representing Human-Robot- Interaction as a dynamical system, and (b) using model predictive control to generate control policies is presented. The approach is motivated by using a scenario in which an Assistive Robot must be productive by bringing work to the human but must also be mindful of the human's workload. By modeling the interaction as a dynamical system, advances in control theory can be leveraged to generate useful control policies.Show more Item Statistical clustering of data(2015-05) Zhang, Lihao; Sager, Thomas W.; Hersh, MatthewShow more Cluster analysis aims at segmenting objects into groups with similar members and, therefore helps to discover distribution of properties and correlations in large datasets. Data clustering has been widely studied as it arises in many domains in marketing, engineering, and social sciences. Especially, the occurrence of transactional and experimental datasets in large scale in recent years significantly increased the necessity of clustering techniques to reduce the size of the existing objects, to achieve a better knowledge of the data. This report introduced fundamental concepts related to cluster analysis, addressed the similarity and dissimilarity measurements for cluster definition, and clarified three major clustering algorithms-hierarchical clustering, K-means clustering and Gaussian mixture model fitted by Expectation-Maximization (EM) algorithm-theoretically and experimentally to illustrate the process of clustering. Finally, methods of determining the number of clusters and validating the clustering were presented as for clustering evaluation.Show more Item Uncertainty propagation and conjunction assessment for resident space objects(2015-12) Vittaldev, Vivek; Russell, Ryan Paul, 1976-; Erwin, Richard S; Akella, Maruthi R; Bettadpur, Srinivas V; Humphreys, Todd EShow more Presently, the catalog of Resident Space Objects (RSOs) in Earth orbit tracked by the U.S. Space Surveillance Network (SSN) is greater than 21,000 objects. The size of the catalog continues to grow due to an increasing number of launches, improved tracking capabilities, and in some cases, collisions. Simply propagating the states of these RSOs is a computational burden, while additionally propagating the uncertainty distributions of the RSOs and computing collision probabilities increases the computational burden by at least an order of magnitude. Tools are developed that propagate the uncertainty of RSOs with Gaussian initial uncertainty from epoch until a close approach. The number of possible elements in the form of a precomputed library, in a Gaussian Mixture Model (GMM) has been increased and the strategy for multivariate problems has been formalized. The accuracy of a GMM is increased by propagating each element by a Polynomial Chaos Expansion (PCE). Both techniques reduce the number of function evaluations required for uncertainty propagation and result in a sliding scale where accuracy can be improved at the cost of increased computation time. A parallel implementation of the accurate benchmark Monte Carlo (MC) technique has been developed on the Graphics Processing Unit (GPU) that is capable of using samples from any uncertainty propagation technique to compute the collision probability. The GPU MC tool delivers up to two orders of magnitude speedups compared to a serial CPU implementation. Finally, a CPU implementation of the collision probability computations using Cartesian coordinates requires orders of magnitude fewer function evaluations compared to a MC run. Fast computation of the inherent nonlinear growth of the uncertainty distribution in orbital mechanics and accurately computing the collision probability is essential for maintaining a future space catalog and for preventing an uncontrolled growth in the debris population. The uncertainty propagation and collision probability computation methods and algorithms developed here are capable of running on personal workstations and stand to benefit users ranging from national space surveillance agencies to private satellite operators. The developed techniques are also applicable for many general uncertainty quantification and nonlinear estimation problems.Show more