부산대학교 도서관

Tractor accidents, especially due to rollover tilting, are one of the main cause of death in agriculture, and a substantial work has been done trying to reduce this risk. Nowadays, it is possible to provide information to the driver adopting global risk index taking into account the ground conditions, the speed and inclination of the tractor, and instantaneous linear and angular accelerations. In this paper the optimal filtering operation of such data obtained by using an Inertial Measurements Unit (IMU) is investigated, in order to provide reliable information without false positive related to local spikes to the driver, on a time scale smaller than the driver reaction time. Data were collected through a properly developed embedded device consisting in a combination of inertial and geo-localization sensors. Experiments were conducted to establish instantaneous values effect on collected data and if these can lead to an erroneous warning for machine operator. Obtained results prove the need of a filtering stage on inertial data in order to reduce false positive warning. Moreover results show how important is danger highlighting in the tractor maneuverability and how driver behaviour changes with that.