Tolerances and Uncertainty in Robotic Systems

摘要

The ability to be programmed for a wide range of tasks is what differentiates robots from dedicated automation. Consequently, robots can be faced with often-changing requirements and conditions. Conventional application development based on teach programming takes robots out of production and occupies personnel, limiting robots’ effectiveness in these environments. Off-line programming solves these problems, but robot inaccuracy must be compensated by a combination of calibration, compliance, and sensing. This complicates up-front systems engineering and application development, but results in systems that can operate in a wider range of requirements and conditions. Performance can be optimized if application tolerances and process uncertainties are known. If they often change, optimization must be done dynamically. Automating this optimization is a goal of smart manufacturing. With its trend of increasing connectivity between the components of robotic systems both within workcells and to the enterprise, exchanging this information has become more important. This includes tolerance information from design through process planning to production and inspection, and measurement uncertainty from sensors into operations. Standards such as ISO 10303 (STEP), the Quality Information Framework (QIF), the Robot Operating System (ROS), and MT-Connect support this exchange to varying degrees. Examples include the assignment of assembly tasks based on part tolerances and robot capabilities; the automated generation of robot paths with tolerances arising from sensed obstacles; and the optimization of part placement to minimize the effects of position uncertainty. This paper examines requirements for exchanging tolerance and uncertainty in robotics applications, identifies how these requirements are being met by existing standards, and suggests improvements to enable more automated information exchange.