Kalman Filter Based Estimation Algorithm to Improve the Accuracy of Automobile GPS Navigation Solution 2014-01-0268
The demand for location based services in automobile industry promoting applications in the area of telematics, vehicle to vehicle and infrastructure communications is encouraging research in the field of accurate navigation solutions. According to the ABI research, position data is the prime enabler for above mentioned applications and the in-car navigation market growth is predicted to grow at 25.9% over the next five years. Consequently position, velocity and heading form the prime input state vector for the target applications. Global Positioning System is one of the accurate off-board sensors for navigation solution. Nevertheless, the cost and complexity of these systems are posing the biggest challenge to the automobile research engineers. Least squares estimation is one of the proven methods used for computing positioning solution under static conditions. However, its accuracy is considered to be poor in dynamic cases.
This work focuses on achieving accuracy within few meters for navigation solution using kinematic based position estimation algorithm. Real-time GPS raw data acquired by tracing a trajectory is further analyzed in a post-mission processing approach. Two positioning estimation algorithms are developed and tested for accuracy and precision. User trajectory obtained using recursive least squares resulted in poor accuracy and bias of about 10m-15m. Subsequently, Kalman filter algorithm was developed to improve the overall accuracy of the kinematic system. The user trajectory for Kalman filter based algorithm was found to be smoother resulting in accuracy within 5m with rare occasional outliers during GPS outages.