Error Correction of Logistical Data: Using a Distributed Viterbi Approach

Student thesis: Master thesis (including HD thesis)

  • Tom Nørgaard Jensen
  • Mads Philipsen
The report deals with a distribution system, where several assets circulate between multiple parties. The assets are registered when being received or returned by one of the parties in the system. This registration is done using RFID technology, which means that the registrations are prone to error. Since the parties in the system are obliged to pay deposit for the assets, it is important that the deposits are calculated on the correct basis. Therefor, it is desirable to correct the data, such that the whereabouts of the assets can be determined even if some registrations are missing.\\ For this purpose an estimator, which can estimate the location sequence of the assets based on the erroneous data, has been developed. The estimator is based on the statistical properties of the distribution system, which can be modelled as a hidden Markov chain.\\ Furthermore, a simulation model of the distribution system has been constructed, in order to test if the estimator is able to determine the correct location sequence.\\ The estimator has been tested in a small scale system. The results have been extrapolated to cover a full scale system. The computational time for the large scale system, show that the algorithm is only usable when distributed between multiple processors. Software for a distributed setup consisting of eleven locations has been designed and implemented in C++. Test results indicate that the estimator in its current condition is not able to fulfil the requirement to its error percentage.
LanguageEnglish
Publication date2008
Number of pages79
Publishing institutionAalborg Universitet
ID: 14420918