Multiple Model Framework of Extended Kalman Filtering for Predicting Vehicle Location using Latest Global Positioning System

This paper looks into various practical methods of improving existing collision avoidance systems that loose accuracy in non-straight paths. Different frameworks of models are studied to identify their accuracy in predicting a vehicle’s future location in non-straight path. From a simple mathematical approach to more complex frameworks of multiple Kalman Filters are evaluated to obtain more accurate collision warnings in situations that current systems can not handle correctly. To make any improvements cost effective, only a Global Positioning System (GPS) receiver was used to obtain all spatial information related to the vehicle. This paper goes through the implementation of several Extended Kalman Filters (EKF) to achieve full coverage of the possible vehicle movements. It then goes on to create multiple model frameworks that combine the different EKF to provide even more accurate estimations of a vehicle’s trajectory. The two frameworks studied in this paper are the Multiple Models Adaptive Estimation System (MMAE) and the Interacting Multiple Models Estimation (IMME). Both the MMAE and the IMME have proven very reliable in the robotics field, so they should also work in the field of automobiles. The last improvement this paper investigates is the integration of the Geographic Information System (GIS) into this system. Road information is available in the GIS data, and spatial information is available through the GPS receiver, therefore, this paper investigates a way to verify that the predicted vehicle’s location is on actual roads. This innovative method reduces the error by more than half when used as part of the system.