Contact Information

RFID-based Part Tracking

Motivation

Customers are increasingly buying goods over the Internet, but still expect to be able to mix and match components as they did at a store. An example would be the purchase of a new computer where there are a number of parameters (memory, hard disk, video card) for the customer to select. This sort of customised manufacture is often referred to as late-stage customisation since all of the different options that the customisation provides can be handled during the last phases of the manufacturing process.

Given this trend toward more flexible production processes where hundreds of different types of end-product are produced by combining component parts in different ways, any automation will need to be more sophisticated. Without having separate lines for each product type, and assuming that the end-product is produced to order rather than to stock, there is the need to rapidly switch between one sort of operation and another. Implicitly, such flexible machines must be able to quickly determine what operation to perform. At least some automation of such built-to-order production is certainly achievable. At Dell Computer's OptiPlex plant, for example, the process of transporting parts around the factory is automated in such a way that each workstation receives only the parts it needs when it needs them. However the final product assembly is still a manual process.

Completely automating such late-stage customisation requires more intelligent automation and better sensory information than have traditionally been available. This is because the decision making in a customisable process does not depend on the mere presence of the part, but on which type of part, and sometimes on the specific identity of that part. For example, computer chassis A will be shipped to customer X , who requires 256Mb of memory, while computer chassis B will be shipped to Y , who requires 1Gb of memory. Thus when a computer chassis arrives at a workstation where memory chips are inserted, A must be treated differently to B . The general problem of establishing and keeping track of the identity and location of physical objects is referred to here as the tracking problem.

Research Aim

It might seem to the casual observer that RFID technology in its current form will solve the tracking problem. However this is not necessarily the case. Some of the hurdles that must be overcome include:

1. An RFID reader does not identify where in its range an RFID tag was when it was detected. Therefore the reader's antenna and / or environment must be configured to reduce the range of possible places where a tag can be detected from. This tends to leave many areas unmonitored by RFID readers, and tags, and therefore tagged parts, will spend at least some time where their location is not being sensed directly in any way. Can we maintain an idea of where parts are and integrate this correctly with subsequent RFID tag read information?

2. Where a large number of RFID readers work in close proximity, they may both tend to read the same tag. This may lead to some confusion about where the tagged part is. This is particularly important in automated systems where the location of the part needs to be detected accurately. Are we able to disambiguate joint tag reads?

3. The RFID reader may be moving. In some cases, it may be possible to integrate information from location sensors, such as GPS, and thus determine the absolute position of a part. Alternatively, known, static tags may provide location information. Is it possible to identify location for a fast moving RFID tag reader?

4. RFID tag readers may completely miss some tags associated with a group of parts (or aggregate) as they move past the tag reader. Such missed tag reads may be inferred based on the aggregate staying together. Artificial Intelligence techniques such as Bayesian reasoning approaches may be employed to perform this inference. Can we detect and deal with missed tag reads using state-based approaches, and if so, what level of protection or improvement do they provide?

The aim of this research is to provide algorithms and theoretical foundations for building RFID-based tracking systems.

Learn more..

For more information about this project, contact James Brusey (jpb54[at]cam.ac.uk)