Proceedings PaperMethod for robot path adaptation using scalar sensor data
|Format||Member Price||Non-Member Price|
An algorithm for modifying a stored nominal robot trajectory path is presented. This method uses data from an arbitrary collection of " one-dimensional" sensors such a sensor returns a scalar position along a fixed vector in space. Examples of such sensors are fixed or robot-mounted proximity probes linear contact probes robotic touch sensing and CCD cameras in conjunction with structured light. The method is not suitable for real-time guidance all the data must be provided at once. The method reassigns the locations of the taught points of the path the spatial resolution of the taught points and the sensing locations must be sufficiently fine to capture any expected shape variation. No physical model of the part or process is required allowing simple and rapid application. The action of this algorithm is quite robust and allows for noise reduction through averaging ofall sensor data. Two heuristic assumptions guide the use of this method: taught points are more affected by nearby sensors than those further away and taught points are moved along the general direction of the sensor vectors. Several parameters allow adjustment of the adaptive action. The algorithm uses a closed-form calculation.