The integration between Global Positioning System (GPS) and MEMS-based inertial sensors provides low cost navigation solutions for many applications. Particle Filter (PF) has been recently introduced as promising non-linear non-Gaussian INS/GPS integration technique that has many merits over traditional linear techniques such as Kalman Filtering (KF) or Extended Kalman Filtering (EKF). However, due to costly computation requirements of PF, real-time implementation is challenging especially if an embedded implementation on a limited resources CPU is targeted. Although the mixture version of PF reduces the computation complexity by reducing the number of particles needed for implementation, further optimization is still needed for embedded systems implementation. In this research, an optimized version of mixture PF is introduced. The optimization is based on further reduction of the number of particles used in PF weighting steps without affecting the accuracy. This reduction is achieved by a fast clustering method called Fast Median Cut Clustering. The PF weighting step is performed only on the representative particles which are the particles cluster centers. The proposed optimized version of mixture PF was implemented on an embedded system utilizing 600 MHz ARM Cortex A8 CPU. The implementation was tested with a Reduced Inertial Sensor System (RISS) integrated with GPS to provide a reliable 3-D land vehicle navigation solution. The proposed real-time system was tested on a mobile robot in real experiments showing fast, robust, and accurate performance. The experimental results show that the computation complexity is reduced by 80% without affecting the accuracy of the overall system. In addition, to the authors' knowledge, this is the first embedded real-time implementation of such computationally costly probabilistic algorithm for INS/GPS navigation.

Additional Metadata
Keywords Embedded real-time systems, Integrated INS/GPS navigation, Land vehicle navigation, Particle Filter (PF)
Conference Institute of Navigation - International Technical Meeting 2011, ITM 2011
Citation
Atia, M, Georgy, J. (Jacques), Korenberg, M. (Michael), & Noureldin, A. (Aboelmagd). (2011). Embedded real-time implementation of mixture particle filter for 3D RISS/GPS integrated navigation solution. In Institute of Navigation - International Technical Meeting 2011, ITM 2011 (pp. 850–857).