Context Detection, Categorization and Connectivity for Advanced Adaptive Integrated Navigation

P.D. Groves, H. Martin, K. Voutsis, D. Walter, L. Wang

Abstract: Context is the environment that a navigation system operates in and the behaviour of its host vehicle or user. The type and quality of signals and environmental features available for positioning varies with the environment. For example, GNSS provides high-quality positioning in open environments, low-quality positioning in dense urban environments and no solution at all deep indoors. The behaviour of the host vehicle (or pedestrian) is also important. For example, pedestrian, car and train navigation all require different map-matching techniques, different motion constraints to limit inertial navigation error growth, and different dynamic models in a navigation filter [1]. A navigation system design should therefore be matched to its context. However, the context can change, particularly for devices, such as smartphones, which move between indoor and outdoor environments and can be stationary, on a pedestrian, or in a vehicle. For best performance, a navigation system should therefore be able to detect its operating context and adapt accordingly; this is context-adaptive positioning [1]. Previous work on context-adaptive navigation and positioning has focused on individual subsystems. For example, there has been substantial research into determining the motion type and sensor location for pedestrian dead reckoning using step detection [2-4]. Researchers have also begun to investigate context-adaptive (or cognitive) GNSS [5-7]. However, this paper considers context adaptation across an integrated navigation system as a whole. The paper addresses three aspects of context-adaptive integrated navigation: context detection, context categorization and context connectivity. It presents experimental results showing how GNSS C/N0 measurements, frequency-domain MEMS inertial sensor measurements and Wi-Fi signal availability could be used to detect both the environmental and behavioural contexts. It then looks at how context information could be shared across the different components of an integrated navigation system. Finally, the concept of context connectivity is introduced to improve the reliability of context detection. GNSS C/N0 measurement distributions, obtained using a smartphone, and Wi-Fi reception data collected over a range of indoor, urban and open environments will be compared to identify suitable features from which the environmental context may be derived. In an open environment, strong GNSS signals will be received from all directions. In an urban environment, fewer strong signals will be received and only from certain directions. Inside a building, nearly all GNSS signals will be much weaker than outside. Wi-Fi signals essentially vary with the environment in the opposite way to GNSS. Indoors, more access points (APs) can be received at higher signal strengths and there is greater variation in RSS. In urban environments, large numbers of APs can still be received, but at lower signal strengths [6]. Finally, in open environments, few APs, if any, will be received. Behavioural context is studied using an IMU. Although an Xsens MEMS IMU is used in this study, smartphone inertial sensors are also suitable. Pedestrian, car and train data has been collected under a range of different motion types and will be compared to identify context-dependent features. Early indications are that, as well as detecting motion, it is also possible to distinguish nominally-stationary IMUs that are placed in a car, on a person or on a table from the frequency spectra of the sensor measurements. The exchange of context information between subsystems in an integrated navigation system requires agreement on the definitions of those contexts. As different subsystems are often supplied by different organisations, it is desirable to standardize the context definitions across the whole navigation and positioning community. This paper therefore proposes a framework upon which a “context dictionary” could be constructed. Environmental and behavioural contexts are categorized separately and a hierarchy of attributes is proposed to enable some subsystems to work with highly specific context categories and others to work with broader categories. Finally, the concept of context connectivity is introduced. This is analogous to the road link connectivity used in map matching [8]. As context detection involves the matching of measurement data to stored context profiles, there will always be occurrences of false or ambiguous context identification. However, these may be minimized by using the fact that it is only practical to transition directly between certain pairs of contexts. For example, it is not normally possible to move directly from an airborne to an indoor environment as an aircraft must land first. Thus, the air and land contexts are connected, as are the land and indoor contexts, but the air and indoor contexts are not. Thus, by only permitting contexts that are connected to the previous context, false and ambiguous context detection is reduced. Robustness may be further enhanced by considering location-dependent connectivity. For example, people normally board and leave trains at stations and fixed-wing aircraft typically require an airstrip to take off and land. The authors would like to acknowledge funding from UK’s Engineering and Physical Sciences Research Council (EPSRC), UCL Engineering, the Chinese Scholarship Council, BAE Systems, Spirent and Terrafix. References [1] Groves, P. D., Principles of GNSS, inertial, and multi-sensor integrated navigation systems, Second Edition, Artech House, 2013. [2] Park, C. G., et al., “Adaptive Step Length Estimation with Awareness of Sensor Equipped Location for PNS,” Proc. ION GNSS 2007. [3] Frank, K., et al., “Reliable Real-Time Recognition of Motion Related Human Activities Using MEMS Inertial Sensors,” Proc. ION GNSS 2010. [4] Pei, L., et al., “Using Motion-Awareness for the 3D Indoor Personal Navigation on a Smartphone,” Proc. ION GNSS 2011. [5] Lin, T., C. O’Driscoll, and G. Lachapelle, “Development of a Context-Aware Vector-Based High-Sensitivity GNSS Software Receiver,” Proc. ION ITM 2011. [6] Shafiee, M., K., O’Keefe, and G. Lachapelle, “Context-aware Adaptive Extended Kalman Filtering Using Wi-Fi Signals for GPS Navigation,” Proc. ION GNSS 2011. [7] Shivaramaiah, N. C., and A. G. Dempster, “Cognitive GNSS Receiver Design: Concept and Challenges,” Proc. ION GNSS 2011. [8] Quddus, M. A., High Integrity Map Matching Algorithms for Advanced Transport Telematics Applications, PhD Thesis, Imperial College London, 2006.
Published in: Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013)
September 16 - 20, 2013
Nashville Convention Center, Nashville, Tennessee
Nashville, TN
Pages: 1039 - 1056
Cite this article: Groves, P.D., Martin, H., Voutsis, K., Walter, D., Wang, L., "Context Detection, Categorization and Connectivity for Advanced Adaptive Integrated Navigation," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 1039-1056.
Full Paper: ION Members/Non-Members: 1 Download Credit
Sign In