Object localisation through clustering unreliable ultrasonic range sensors

The increasing popularity and availability of inexpensive ultrasonic sensors facilitates opportunities for tracking moving objects by using a cluster of these sensors. In this paper, we use a cluster of ultrasonic sensors connected to Raspberry Pis. Field tests indicate that the accuracy and reliability of individual sensors depends on the relative position of the tracked object. Hence, we employ data fusion and synchronisation techniques for trilateration to improve accuracy using a cluster of sensor nodes. We successfully conduct multiple runs tracking a moving object and report these field test results in this paper. Our average error is in the order of tens of centimetres, and some of our best results match published results for larger clusters.