A new non-linear filtering algorithm for road-constrained vehicle tracking

Road constrained vehicle applications such as Intelligent Transport Systems (ITS) and Location Based Services (LBS) have become much more widespread over the last decade, creating the need for effective solutions to the problem of reliable and accurate road-constrained vehicle positioning. While the problem of tracking has been to a satisfactory degree solved for some applications in good GNSS visibility situations, this is not the case where satellite signal quality is degraded or non-existent particularly where the application is reliant on ubiquitous high quality positioning. Attention has increased over the last decade on formulating signal outage handling algorithms however we argue that the problem is far from comprehensively solved. Such outages can still cause significant disruption to positioning accuracy even when occurring over short period. We argue that in principle the problem of bridging partial outages (between 1 and 3 satellites visible, inclusive) can be adequately solved when accurate digital road map data is combined ("fused") effectively with partial satellite information in a statistically robust way. Our contribution is a statistically rigorous positioning algorithm which implicitly fuses road map data with satellite range measurements to sequentially estimate the position of a moving on-road vehicle. An innovative approximation scheme to handle network non-linearities efficiently is incorporated using Gaussian sums and locally linear models. We present results showing the effectiveness of this algorithm when compared to benchmark methods such as the Extended Kaiman Filter (EKF) and map-matching. Tests involved error comparison of tracking accuracy between algorithms over periods of (artificially induced) partial signal outages.