We develop convergent minimization algorithms for Bethe variational approximations which explicitly constrain marginal estimates to families of valid distributions. While existing message passing algorithms define fix...
详细信息
ISBN:
(纸本)9781627480031
We develop convergent minimization algorithms for Bethe variational approximations which explicitly constrain marginal estimates to families of valid distributions. While existing message passing algorithms define fixed point iterations corresponding to stationary points of the Bethe free energy, their greedy dynamics do not distinguish between local minima and maxima, and can fail to converge. For continuous estimation problems, this instability is linked to the creation of invalid marginal estimates, such as Gaussians with negative variance. Conversely, our approach leverages multiplier methods with well-understood convergence properties, and uses bound projection methods to ensure that marginal approximations are valid at all iterations. We derive general algorithms for discrete and Gaussian pairwise Markov random fields, showing improvements over standard loopy belief propagation. We also apply our method to a hybrid model with both discrete and continuous variables, showing improvements over expectation propagation.
In order for robots to interact safely and intelligently with their environment they must be able to reliably estimate and localize external contacts. This paper introduces CPF, the Contact Particle Filter, which is a...
详细信息
ISBN:
(纸本)9781509037636
In order for robots to interact safely and intelligently with their environment they must be able to reliably estimate and localize external contacts. This paper introduces CPF, the Contact Particle Filter, which is a general algorithm for detecting and localizing external contacts on rigid body robots without the need for external sensing. CPF finds external contact points that best explain the observed external joint torque, and returns sensible estimates even when the external torque measurement is corrupted with noise. We demonstrate the capability of the CPF to track multiple external contacts on a simulated Atlas robot, and compare our work to existing approaches.
Path finding is an important problem in robot design and automation that requires quick error-free solutions that rely on external environment. Automated mobile robotic systems employ various techniques to determine t...
详细信息
ISBN:
(纸本)9781538674765
Path finding is an important problem in robot design and automation that requires quick error-free solutions that rely on external environment. Automated mobile robotic systems employ various techniques to determine the path that the robot needs to follow to reach the destination to perform its function. Path finding problems can utilize various algorithms to solve the problem. Sensor data can be used as a reference to determine the path to be followed from the start point to the destination. However, this technique is highly localized and does not provide the ability to make decisions by taking global constraints or conditions into consideration. Image processing techniques are employed extensively to provide a solution based on global conditions. The proposed method involves use of image processing to process the acquired image of the maze from a mounted camera system. The processing steps are used to provide steps which the robot can follow to reach from its current position to the final position. This project implements a universal algorithm to allow the robot to maneuver autonomously.
Simulation models typically describe complicated systems with no closed-form analytic expression. To optimize these complex models, general "black-box" optimization techniques must be used. To confront compu...
详细信息
ISBN:
(纸本)9781467397414
Simulation models typically describe complicated systems with no closed-form analytic expression. To optimize these complex models, general "black-box" optimization techniques must be used. To confront computational limitations, Optimal Computational Budget Allocation (OCBA) algorithms have been developed in order to arrive at the best solution relative to a finite amount of resources primarily for a finite design space. In this paper we extend the OCBA methodology for partition based random search on a continuous domain using a lookahead approximation on the probability of correct selection. The algorithm uses the approximation to determine the order of dimensional-search and a stopping criterion for each dimension. The numerical experiments indicate that the lookahead OCBA algorithm improves the allocation of computational budget on asymmetrical functions while preserving asymptotic performance of the general algorithm.
暂无评论