AmbiSense: Identifying and Locating Objects with Ambient Sensors
In order to simplify processes in logistics, warehousing, and surveillance, our interdisciplinary joint project AmbiSense has combined solutions for efficient acquisition and mapping of environments. These environments are equipped with diverse ambient technology such as WLAN, Bluetooth, and RFID. The research is based on techniques stemming from the fields of robotics, embedded systems, augmented reality (AR) and Enterprise Resource Planning (ERP).
More precisely, we present a novel complete system for machine-aided inventory. Our system covers automatic product identification using RFID, localization based on ambient sensors, the enrichment of raw RFID data with product information from ERP backend systems and real-time augmented reality visualization.
One key component of our project is the continuous integration of all developed algorithms and techniques into a real-world demonstrator to illustrate their practicability and usefulness. We have chosen warehousing and retail as our current application scenario: Robot-assisted inventory is applied in a supermarket as we expect goods to be labeled individually with RFID tags in the near future. This enables products to be tracked from production to sale consistently and to be localized permanently.
In order to provide a working demonstrator, we set up an application scenario resembling a supermarket at the AmbiSense lab at the University of Tübingen. It consists of individually tagged products placed in typical shop shelves. Our robot, equipped with an RFID reader, traverses the supermarket environment while constantly detecting products within its range. The data are transmitted using WLAN to a central computer which holds a model of the current state of the system. We augment these data by additional product-specific information provided by the ERP system. The detected objects as well as additional product data are visualized using AR techniques.
This scenario aims at synchronizing the product stock of supermarkets or stores automatically. Other sample tasks could be the identification of products that are past their sell-by dates or located in the wrong places.
In addition the robot localizes itself using the existing infrastructure of different, cost-efficient ambient wireless sensors. To achieve the location we develop and combine novel positioning techniques using passive UHF RFID, Bluetooth, and WLAN. We thereby employ three orthogonal measuring techniques: detection rates, signal strength, and round trip time. The orthogonality of the methods is designed to achieve robustness to noise and unforeseen changes in the surroundings. Moreover, due to their different read ranges, the technologies can complement each other at different scales of the environment. An effective and cost-efficient indoor location solution can only be achieved with multiple and heterogeneous ambient sensors combined together.