Sensor data fusion using perception net for a precise assembly task

In this paper, work on sensor fusion technology is presented for a peg-in-hole insertion task. Three kinds of sensors are fused for the task execution. They are a vision sensor, a proximity sensor, and a force/torque sensor. The fusion technique is based on the perception net. The perception net, as a structural representation of the sensing capabilities of a system, connects features of various levels of abstraction, referred to here as logical sensors, with their functional relationships such as feature transformations, data fusions, and constraints to be satisfied. The vision and proximity sensors are mainly used for the gross motion control and the force/torque sensor is additionally used for the fine motion control of the robot. The perception net is introduced to exploit the merit of each sensor with putting weight on the sensor. The effectiveness of the proposed sensor data fusion method is validated through simulations