A practical method for improving mobile devices trajectory estimation via MEMs accelerometers

Abstract

The challenging research area of locating a mobile device has gained interest in recent years with the emergence of wireless sensor networks (WSN) as well as of low-cost sensors based on micro-electromechanical systems (MEMS). This paper focuses on the dead reckoning (DR) algorithm which doubly integrates the specific force sensed by MEMs acclelerometers to increment the current position of the mobile device. As the classical DR algorithm is very sensitive to MEMS accelerometers noise, a practical method, named the MDR method — for Modified DR-, that improves the accuracy of the classical DR algorithm is proposed. It provides inference on the presence/absence of movement, and uses this inference to mitigate the impact of accelerometer noise. Performance improvement of the MDR algorithm over the classical one has been demonstrated, both in simulation and in a real test environment where an industrial Kuka arm robot has been used to generate a known and controlled trajectory to a Crazy Flie quadcopter which serves as the mobile device. Mean absolute errors (MAE) less than 3 centimeters are obtained by the proposed MDR, when the classical DR attains MAE of the order of 15 centimeters.