Sensor Localization

Sensor-Assisted Localization

Sept. 2003 ~ Oct. 2004

Ubicomp Lab, National Taiwan University

Advisor: Prof. Hao-hau Chu


Wi-Fi based indoor location systems have been shown to be both cost-effective and accurate, since they can attain meter-level positioning accuracy by using existing Wi-Fi infrastructure in the environment. However, two major technical challenges persist for current Wi-Fi based location systems, instability in positioning accuracy due to changing environmental dynamics, and the need for manual offline calibration during site survey. To address these two challenges, three environmental factors (people, doors, and humidity) that can interfere with radio signals and cause positioning inaccuracy are identified. Then, we have proposed a sensor-assisted adaptation method that employs RFID sensors and environment sensors to adapt the location systems automatically to the changing environmental dynamics. The proposed adaptation method performs online calibration to build multiple contextaware radio maps under various environmental conditions. Experiments were performed on the sensor-assisted adaptation method. The experimental results show that the proposed adaptive method can avoid adverse reduction in positioning accuracy under changing environmental dynamics.

A Glance

Figure 1 shows our scenario and experiment testbed. The red triangles represent the WiFi Access Points and the blue circles represent RFID readers. We know the location of RFID readers in advance and use them as the landmarks. When the users are close to these landmarks, we learn their locations. And then we assume the users walk in constant speed between landmarks as long as the inter-arrival time between two landmarks is reasonable. By interpolation, we can learn the coarse locations of the clients by just using these landmarks.


Figure 1. Testbed.

To train the radio map for the WiFi based localization system, we use the coarse locations and their timestamps to find the corresponding RSSI collected by the clients as they are walking. Although the locations learned from interpolation are not very accurate, but when the number of traces used for training increases, we can learn better radio map as shown in Fig 2. And Fig. 3 shows that in 90% the prediction error is smaller than 5 meters when we collect enough training data.

#Traces and Performance

Figure 2. when the # of training traces increases, the prediction error decreases.


Figure 3. Evaluate our system in both indoor and outdoor environment.


Yi-Chao Chen, Ji-Rung Chiang, Hao-Hua Chu, Polly Huang, Arvin Wen Tsui: Sensor-assisted wi-fi indoor location system for adapting to environmental dynamics. MSWiM 2005: 118-125