Location sensing & routing protocols

Sensing and tracking the location of slowly moving objects in indoor environments has considerable potential for enabling novel location-aware services and applications. In the retail industry, for example, a shopping cart equipped with a personal shopping assistant and location-sensing functionality can guide customers through the store, deliver location-based product information, and alert the customer to promotions and discounts (Fig. 1).  

We are working on the design and development of a novel solution for locating and tracking shopping devices in a retail store. The system uses a combination of a wireless reference positioning system (WRPS) and an inertial navigation system (INS) (Fig. 2). The WRPS provides location estimates based on radio signal-strength measurements and triangulation with at least four radio beacons sent from known positions. The estimates have a precision of a few meters and good long-term stability.  By averaging the radio signals over both time and space using a multi-antenna radio receiver, and by taking into account spatial constraints given by the floor plan of the building, the accuracy can be improved further. The sensor-based INS system measures the acceleration and angular velocity in a given coordinate system and produces high-precision, short-term position estimates. However, drift inherent in an INS system limits long-term stability. We are therefore pursuing a combination of WRPS and INS where an extended Kalman-filter algorithm combines the short-term estimates produced by the INS with the long-term estimates obtained from the WRPS. A prototype sensing and tracking system has been built that achieves an accuracy of 1-2 m with high reliability (Fig. 3).

Consider a wireless sensor and actuator network (WS&AN) comprising a large number of battery-powered, computationally limited sensor nodes deployed at fixed locations, and a few computationally more powerful actuator nodes that move around in the sensor field (Fig. 4). Multi-hop transmission is used to route the sensor data. In such a scenario, the sensor and actuator nodes must cooperate to form an ad-hoc network, which can forward data efficiently in a dynamically changing environment, minimize power consumption, and maximize network lifetime. Therefore, scalability, energy efficiency, and the support of node mobility are the primary challenges for the design of such a WS&AN.

We are working on geographic routing concepts where each sensor node selects the next-hop forwarding node based on its own geographical location, the position of its neighbors, and the position of the destination node. As the next-hop node selection algorithm requires only a local information exchange, the algorithm scales well for a network with a large number of sensors. Energy efficiency is achieved with a cross-layer approach where protocols, which are assigned to separate layers in the classical communication model, are jointly designed and optimized to reduce energy consumption.

Support of node mobility in the WS&AN is obtained by incorporating a Kalman filter into the mobile nodes for sensing their geographic location and transfer the sensed location information with a power-efficient dissemination protocol to static source nodes. By exploiting this knowledge, the source nodes can predict the location of the mobile nodes in order to efficiently route data packets through the sensor network to the mobile destinations using geographic routing.

Figure 1. Retail store of the future.

Figure 2. Combined location-sensing system.

Figure 3. Experimental results.

Figure 4. Geographical routing.