1. Position Detection and Identification of Products using RFID Technology Master Thesis Christian Decker Supervisor: Michael Beigl This thesis representation was held at TecO on September 20 th 2002 http://www.teco.edu/~cdecker/pub/publications.html

2. Structure Problem + Today‘s Situation Goal SmartShelf Technology Prototypes Conclusion

3. Problem quantity+type of products (purchase) quantity+type of products (sale) Scenario: traditional retail business Quantity of products on shelfes (Out-of-Stock problem) History of products Success of product placement ???

4. Today‘s Situation Refill of shelfes happens at fixed points in time – not driven by necessity Product placement rearrangement of products in stores to stimulate the consumer‘s buying behavior (Surprise-Effect) test of sales combinations of similiar products (trial-and-error) There is no quantitative proposition about the interactions of the consumer!

5. Goal: Interaction Detection System Consumer buying behavior remove, add, move (basic interaction patterns) Improvement of existing systems automatic re-order systems statistics Provide new services recommender-, broker-, help desk-systems electronic pricing

6. SmartShelf – ID Technology Identification and position detection of products using RFID insensitive to dirt, wetness etc., tiny, contactless, no internal power supply products equipped with transponders (price!) H400x EM Marin 125kHz not collision aware 130ms read cycle 40bit ID,read-only

7. SmartShelf – Design Identification every transponder has its own unique ID Position detection – 3 reading units in parallel big detection surface divided into smaller RFID detection spots (here: only 1 transponder per spot)

8. SmartShelf – Antennas Requirements local detection through “limited” reading field homogenous reading field big detection surface with only a few antennas

9. SmartShelf – Function Detection – Gathering – Communication

10. SmartShelf – Sychronisation Synchronisation of communication serial communication, TTL, fixed length of data among the reading units: start of communication is determined via signals (READY,Request(RQ) ) to external systems: CTS signal and Command/Response Sychronisation of antenna activation fixed schema for activation of antennas avoids reading collisions of adjacent detection spots avoids simultaneous activation of two antennas on the same reading unit

11. SmartShelf – Synchronisation 2 Timing synchronisation max. time for activation+reading of an antenna is taken to determine the start of the next step in the program’s execution Address sychronisation central unit addresses an antenna for reading, sets up the time pattern for all Barrier synchronisation all reading units work independently until a sync point – program blocks there! global signal to continue the program’s execution

12. SmartShelf – Software Memory function problem: assignment of transponders is not stable assumption: detected transponders do not disappear suddenly if apparently disappeared, then several read trials increases stability, but needs time Fast Update transmission of data starts even if not all antennas were checked again for transponders disadvantage: part of the informationen is out of date

13. SmartShelf – Prototypes 1 st Prototype: built completely by hand address sychronisation error-prone hardware, instable software, slow 2 nd Prototype: PCB, optimization of antenna’s characteristic barrier synchronisation, memory function, fast update hardware reliable, fast, 99.7% detection reliability 3 rd Prototype: spike filter tweaking of parameters without reprogramming fast, automatic tweak control

14. Conclusion Identification+position detection (99.7% reliable) Basic interaction patterns (remove, add und move) can be detected – consumer behavior can be acquired quantitativly Detection speed too low – Goal: realtime detection of interaction patterns Standard interface allows integration in existing systems and development of new applications (recommender and broker systems)