A tripartite filter design for seamless pedestrian navigation using recursive 2-means clustering and Tukey update

Mobile devices are desired to guide users seamlessly to diverse destinations indoors and outdoors. The positioning fixing subsystems often provide poor quality measurements with gaps in an urban environment. No single position fixing technology works continuously. Many sensor fusion variations have been previously trialed to overcome this challenge, including the particle filter that is robust and the Kalman filter which is fast. However, a lack exists, of context aware, seamless systems that are able to use the most fit sensors and methods in the correct context. A novel adaptive and modular tripartite navigation filter design is presented to enable seamless navigation. It consists of a sensor subsystem, a context inference and a navigation filter blocks. A foot-mounted inertial measurement unit (IMU), a Global Navigation Satellite System (GNSS) receiver, Bluetooth Low Energy (BLE) and Ultra-wideband (UWB) positioning systems were used in the evaluation implementation of this design. A novel recursive 2-means clustering method was developed to track multiple hypotheses when there are gaps in position fixes. The closest hypothesis to a new position fix is selected when the gap ends. Moreover, when the position fix quality measure is not reliable, a fusion approach using a Tukey-style particle filter measurement update is introduced. Results show the successful operation of the design implementation. The Tukey update improves accuracy by 5% and together with the clustering method the system robustness is enhanced.