Probabilistic reasoning for indoor positioning with sequences of WiFi fingerprints

The paper tackles the problem of indoor personal positioning using sensors available in modern mobile devices, such as smartphones or tablets. Alike many of the state-of-the-art approaches, the proposed method utilizes WiFi fingerprints to find the user's position in a predefined map of WiFi signals. However, we improve the approach to WiFi-based positioning by considering probabilistic dependencies between the neighboring fingerprints in a sequence of consecutive WiFi scans. The algorithm uses linear-chain Conditional Random Fields to infer the most probable sequence of user's positions, which makes it possible to find a consistent trajectory. Due to the use of probabilistic reasoning in a wider spatial context the algorithm considers a number of possible positions, and resolves ambiguities stemming from noisy WiFi measurements. We tested the approach using data collected in one of the buildings of Poznan University of Technology with a regular smartphone.