METHOD AND SYSTEM FOR TRACKING MULTIPLE OBJECTS

Patent №

US 6,704,692

Granted

2004-03-09

Filed 2000

Owner

BOEING COMPANY, THE

Lab

AI components

5

ml · vision · planning · evo · hardware

Assignment

Recorded

Dataset

AIPD

2023_r1 edition

Application

09696473

An improved method and system for solving a combinatorial optimization problem, such as a tracking problem, to define a plurality of associations of measurements taken of a plurality of objects is provided. In one aspect, a method, a system and a computer program product are provided for constructing a plurality of updated tracks by solving a Lagrangian dual in which each of the measurement constraints has been relaxed. In another aspect, a hybrid branch and bound and LR technique is provided to select a plurality of updated tracks from among a plurality of candidate tracks having respective initial costs. In this regard, a search tree of the candidate tracks is ordered based upon the initial costs of the candidate tracks as adjusted by the dual variables that have been defined as a result of solving a Lagrangian dual. In order to attempt to increase the efficiency with which a Lagrangian dual is solved by nonsmooth optimization techniques, initial values for the dual variables and some of the subgradients are judiciously selected. The dual variables are initialized and some subgradients are provided based upon values of corresponding dual variables and some of the subgradients, respectively, that were determined during the solution of the prior problem. The method and system can be implemented in a parallel processing architecture that utilizes both coarse grain and fine grain techniques to evenly schedule a number of subproblems amongst a plurality of processors in order to obtain a solution in an efficient manner.

AI classification

Machine learning1.00
Planning1.00
Evolutionary computation1.00
AI hardware0.73
Vision0.59
Knowledge representation0.17
Speech0.00
Natural language0.00

Ownership

BOEING COMPANY, THE

assignment · 116940403

Assignors

BANERJEE, SUBHANKAR, BERGE, MATTHEW ELDEN, DEVUN, ESMOND ERNEST JR., FILIPOWSKI, SHARON KAY

On an employer assignment, the assignors are typically the inventors.

© 2026 NYSGPT2525 LLC