Deliver a suite of tools that facilitates calibration procedures for individual robots, robot-to-robot coordination, sensors, and grippers to mitigate the lack of automation and technical expertise that currently prevents small and medium manufacturers from adopting robotic systems.
What is the Technical Idea?
SMEs could benefit greatly from the use of robots in the factory, but are limited by the technical challenges inherent to installation and use of robots and related sensors and tooling within dynamic unstructured environments. Automating the technically arduous processes that are currently required for calibration of robot arms, sensors, and grippers will reduce these challenges preventing many SMEs from adopting robots. NIST will apply its expertise in metrology for sensor and robot calibration to produce streamlined “recipes” for characterizing sensors, kinematics, and dynamics that can be used by SMEs without requiring deep technical knowledge. The idea is to move the domain from an ad hoc process through a more rigorous framework that sensor and robot developers, systems integrators, and end users can leverage. Just like consumer products, such as gaming systems, that are starting to incorporate sophisticated sensors that don’t require extensive calibration and setup, NIST will produce a tool set that will allow easier adoption of robotic systems within SME shops.
Robot calibration is a procedure that is used to determine the kinematic and dynamic parameters that define the geometric and mechanical properties of a robot structure. Correctly estimating these parameters is essential for improving and maintain robot position and velocity accuracy. In order to establish the parameter values, robot modeling techniques, accurate measurement equipment, and means of reliably identifying the parameter values are needed. Sensors must also undergo calibration procedures to establish their characteristics (e.g. error profile). Furthermore, the correlation between the coordinate frames of various sensors among themselves (also known as “registration”), to achieve fusion, and with the robot arm itself, must also be established to enable effective collaborative and agile systems. The difficulties inherent in performing these measurements and determining the correct parameters are currently daunting for non-experts and the complexities will increase as more and different sensors (e.g, 3D range imagers) and collaborative robots that are not fixed to the shop floor become more common . NIST will provide the measurement science foundations for streamlining all of these steps.
What is the Research Plan?
A workshop, held in 2015 with small and medium manufacturers, helped establish the current state of the practice and enumerate the spectrum of barriers to higher utilization of robotics, especially in assembly-centric operations. The workshop report (https://www.nist.gov/sites/default/files/documents/2017/02/07/smallmfrroboticswspreport-public-nistmep-081816-final.pdf) is helped steer the detailed definition of the research agenda. Additional input is being generated through ongoing discussions with small manufacturers through NIST’s Manufacturing Extension Partnership (MEP). The key areas of focus are: (1) individual robot arm calibration; (2) sensor calibration and registration (sensor to robot coordinate frame and between sensors); (3) multi-robot collaboration; and (4) streamlining offline programming for robotic systems.
NIST built a testbed to instantiate the highest priority calibration challenges expressed by SMEs. The testbed is being used as a resource with which to study the problems, produce reference solution implementations, and verify these solutions. A second testbed is under construction, and is expected to be commissioned in FY19. The research plan focuses on the following areas, which will be presented in a coherent web-based framework to guide application of the tools.
- Calibration: NIST is developing simplified and fast methods for calibrating and registering robot arms with respect to their work surfaces. These methods include a “sensor free” approach for deriving the parameters needed to improve a robot’s kinematic accuracy and a three-point procedure for registering a robot to the coordinate system for its task. Additionally, NIST is developing evaluation techniques that can be used to determine the best approaches for registering a robot to its workspace.
- Sensors: NIST is conducting extensive experiments to develop the mathematical foundations for (a) facilitating calibration of sensors, including multi-camera optical tracking systems, (b) most expediently establishing the transformations from one coordinate system to another, and (c) providing guidance for placement of sensors and/or fiducials used during the registration process to minimize error of the resulting transformation matrix.
- Registration: Mobile manipulators are a form of multi-robot systems that collaborate and must have the arm’s coordinate system registered with that of the vehicle. NIST is developing methods for achieving and evaluating registration of arm to vehicle, mobile manipulator to work area, and for registering two or more robot arms so that they can jointly accomplish a task.
- Modeling: The methods and tools developed for calibrations and registrations will be applicable to streamlining the implementation of robot programs created offline. NIST will investigate incorporating more realistic models of robots, sensors, and the workcell and workpieces within simulations and off-line programming systems to reduce or eliminate the extensive "touching up" currently needed before being run on the real robot.
Ultimately, NIST will produce robot models, datasets, software tools, and calibration artifacts that can lead to easily calibrated or even self-calibrating sensors and robots.