1 2 3 THREE ASPECTS OF THIS PRESENTATION IN-LINE ROBOT CALIBRATION IN ABSOLUTE SPACE NETWORKING CAPABILITY IN ABSOLUTE SPACE IN-LINE QUALITY ASSURANCE IN ABSOLUTE SPACE A presentation by Andreas Demopoulos, MD Absolute Robotics Ltd 18 Marley Fields, Leighton Buzzard Bedfordshire, LU7 4WL
THREE ASPECTS OF THIS PRESENTATIONIN-LINE ROBOT CALIBRATION IN ABSOLUTE SPACENETWORKING CAPABILITY IN ABSOLUTE SPACEIN-LINE QUALITY ASSURANCE IN ABSOLUTE SPACE
A presentation by Andreas Demopoulos, MD
Absolute Robotics Ltd
18 Marley Fields, Leighton Buzzard
Bedfordshire, LU7 4WL
Tel:- 01525 375967
Mobile:- 07751 482131
E-mail :- [email protected]
THE BASE RING
This is a mechanically and thermally stable structure made of NILO or INVAR material. The base ring has three parts:--
a) Magnetically detachable SMRs. After ring manufacture, the centres of SMRs are measured and used to define the local reference system of axes XYZ. SMRs provide connectivity to the outside world via a laser tracker so that local references and work objects can be networked together.
b) Magnetically detachable feet. These provide exact repositioning capability and ring positional verification. The legs are cemented to the ground where robots are fixed or expected to move within the cell.
c) Reference elements. These items depend on the measurement system employed. If the measurement system is a laser tracker those elements are simply the the existing SMRs. If the measurement system is based on photogrammetry, the elements are reflective targets or other geometric shapes. In our case the elements are imaging sensors (no lens). Whatever the elements may be, they must be calibrated so that their positions are known w.r.t the local absolute frame of reference XYZ.
Laser beam generators
THE BEAM GENERATOR
This is a mechanically and thermally stable structure that has two parts :-
a) Magnetically detachable SMRs. These define the local reference axes ABC and provide networking capability to the outside world via a laser tracker.
b) The laser system that generates multiple beams in two orthogonal planes using Diffractive Optical Elements (DOEs.
We have three things:-
a) Two systems of axes XYZ and ABC that can move relative to each other
b) Three non-parallel lines k, l and m that are fixed relative to ABC system of axes and whose vector equations are known with respect to the ABC system of axes. Lines could be co-planar.
c) Three points P1, P2 and P3whose position vectors are known with respect to the XYZ system of axes.
If points P1, P2 and P3 happens to lie on lines k, l and mthen the relative position and orientation of the two systems of axes can be uniquely determined with respect to one another.
The Mathematical Principle
Determining the vector equations of laser beams
A thermally and mechanically stable ring is used to establish the equations of laser beams. This calibration ring has a number of SMRs, the centres of which define the local frame of reference EFG. An imaging sensor is also attached to the calibration ring. The sensor is calibrated so that the pixel positions are known w.r.t EFR system of axes.
The laser beam generator is fixed. A laser tracker locates the SMRs on the laser beam generator. This means that the position of the laser tracker is established relative to the local frame of reference EFG and hence any subsequent measurements the laser tracker makes can be referred back to the EFG frame of reference.
The calibration ring is moved along the axis of each beam in a point-to point fashion and the position of the ring is located by a laser tracker. By sensor calibration, the centre of the light spot is known w.r.t EFG system of axes and via the laser tracker it is related to the ABC system of axes. In this way we obtain the co-ordinates of points along the axis of each beam and thus their equations w.r.t to ABC system of axes.
The calibration process determines the position and orientation of the robot’s axes in absolute space at discrete robot configurations. Intermediate configurations are determined by the axes encoders.
Laser beam generator is attached to the robot’s end effector
The basis procedure is to rotate one axis at a time while keeping all others fixed. As a given axis rotates, different beams engage with different sensors on the base ring and thus the position and orientation of the rotated axis is measured at discrete angular intervals w.r.t to the local frame of reference XYZ. At the same time, the encoder positions of all axes are recorded. This enables interpolation between calibrated positions.
The A6 axis is rotated first and the A4 axis next. It is convenient to have the two axis aligned so that the same sensors are used on the base ring. The A5 axis is rotated next. The other set of laser beams that is orthogonal to the first is used in this case to engage with a different set of imaging sensors on the base ring. The process continues with the A1 axis, and concludes by rotating the A2 and A3 axes. The sensors on base ring are strategically placed so as provide engagement with corresponding laser beams at the discrete robot configurations.
How do we deal with blind spots? First of all, the measurement system is located behind the tools attached to the robot, not in front of them. This must help to avoid blind spots. Secondly, since the robot is not calibrated in a point-to-point fashion, it doesn’t matter where we choose to spin a particular axis provide it we spin it at the same robot configuration. The changes to an axis’s position and orientation are going to be the same. Therefore, we can chose to spin an axis in a visible configuration; It doesn’t have to be inside a blind spot where the laser beam generator cannot view the imaging sensors on the base ring.
The key concept is that we do not network robots directly. This is plainly not possible. Instead, we network local references during initial cell set-up and calibrate robots against the known and verifiable positions of their local references. This is how is done.
Locating legs are buried in the ground at pre-determined locations where robots are fixed or expected to be moved during production.
Base rings are attached to the legs and local references XYZ are related to each other during the initial cell set-up using laser trackers.
Robots are placed within the base rings and receive absolute information from their base rings. When robots move within the cell, they take along their rings and place them onto the 3 locating legs in the new robot position. The base rings are designed to re-locate exactly on the same spot and their positions are verified automatically. Robots are then calibrated against the known and verifiable positions of their base rings
Robots become independent absolute manufacturing and measurement tools that move around the cell as autonomous guided vehicles . Since the locating feet are buried in the ground, the floor is totally clean and uncluttered as no 7th axis is required. This solves the large volume manufacturing problem.
Since robots are absolute working devices, relating robots to their work objects is easily accomplished by attaching various metrology probes on the robot’s flange, thus using effectively the robot as a CMM machine.
Any system of absolute calibration is useless without in-line quality assurance. This is because the calibration process predicts the robot’s position whereas the quality assurance verifies it.
Unless we have some independent means to measure directly the position and orientation of the various tools attached to the robot we do not know if their predicted positions are correct or not. This is why we need to have a QA system alongside an Absolute calibration system.
If the ultimate goal is to achieve zero scrap rate we must avoid mistakes happening at the first place. This means that for each operation we must know where the tools are and what they are doing. Since robots work in absolute space, we can attach conventional manufacturing tools to the robot such as drills and cutters as well as measurement tools such as various stereo and laser scanners. In this way we can accomplish an operation and measure its outcome with the same robot but with different tools. However, adding measurement capability to a robot is not the same as adding QA. QA is to know that our measurements are correct. This is what the in-line Quality assurance does.
Why do we require in-line QA?
THERE ARE TWO LEVELS OF QA.
In the 1st level we check the position and orientation of the tools and their functionality
For the purpose of QA we treat everything as “tools” irrespective of whether they are measurement or manufacturing tools, since all we are interested is if they are in the right place and doing the right things. The specific measurements depend on the function of the tool itself. If a tool is for example a drill all we have to measure is its diameter and absolute position and orientation in space. If the tool is a laser profiler all we have to measure is the absolute position and orientation of their laser planes and the calibration of the cameras.
The purpose of the base ring is now expanded beyond to that of simply providing local absolute references. The base ring incorporates now specific probes that relate to the functionality of the tools attached to the robot. Once installed on the base ring, the probes are calibrated against the local references XYZ so their absolute positions are known. After initial tool set-up, the tools are engaged with their respective QA probes and their initial absolute positions are recorded. During manufacture the tools are re-engaged with their respective QA probes and their positions are compared against their initially recorded positions. Any difference beyond a set limit, stops production and triggers a diagnostic system.
In the 2nd level of QA we check the accuracy of the Absolute system itself
The position of the ABC system of axes at the robot’s flange can be determined by two independent systems; one is the 6-DOF laser beam measuring device and the other is a laser tracker. Therefore the laser tracker can be used independently to check the absolute position of the ABC system of axes.. The laser tracker thus defines the 2nd level of QA and since it is an expensive equipment, the laser tracker is used in three cases:- a) Initial calibration of the absolute system b) periodic inspection and c) as a diagnostic tool.
How it works?
Probes for laser profilers and line sensors
The camera position is determined from Tsais algorithm by viewing a number of non-coplanar holes or photogrammetry targets whose absolute positions are known w.r.t the XYZ system of axes. The absolute position of the laser plane or laser line is determined by intersecting imaging sensors whose pixels are calibrated against the XYZ system of axes.
Probes for axi-symmetrical tools
The robot spins the tool within a cluster of 6 proximity sensors arranged in a star-like configuration in two layers that are spaced a known distance apart. The device determines the tool diameter and the position of its axis. A further sensor determines the tool tip.
Other QA probes
QA probes could be designed for example to check the thickness and orientation of a water jet, by a number of laser micrometers arranged in two planes.
Examples of tool-specific GA probes
There are three aspects that makes this calibration process highly accurate.
No chain errors since we measure directly how the distortion of each articulation arm affects the position and orientation of the robot’s flange.
Laser beams strike directly the imaging sensors so the slightest displacement and rotation of the beam generator and any part to which the generator is attached too, can be measured with very high accuracy, irrespective of the distance. This means that very large robots can be calibrated with the same ease as small ones.
The robot is calibrated over the entire range of its axes not a small part of it.Intermediate positions are obtained by interpolation between the calibrated positions. This is a lot more accurate than other systems that swing the robot a few degrees around a ball and use the robot’s kinematic model to extrapolate outside the calibrated positions, in other words, trying to guess how the robot behaves outside the measured regions.
Temperature and stiffness related errors
TEMPERATURE RELATED ERRORS
Temperature related errors are cycle and environmental conditions dependant and hence, the only way to deal with those errors is to calibrate the robot periodically in-line under its actual working conditions.
STIFFNESS RELATED ERRORS AND THE ACTIVE STIFFNESS CONCEPT
In the active stiffness concept the output from the force-torque sensors attached at the robot flange causes the joints to stiffen up almost instantaneously in the opposite direction to the applied load so that the net deflection is zero. The robot controller works in open loop mode, sending the correct current directly to each servo-motor in accordance to look-up tables that are populated during the stiffness calibration process. Therefore, the controller does not have to wait for positional errors to occur in order to drive the closed loop because the correct currents to each servo motor are already known from the stiffness calibration process.
Since stiffness related errors are not affected in any significant way by temperature and working conditions, the stiffness calibration process can carried out once at the robot factory and is briefly as follows:- For a given robot configuration, a known load is applied along a known direction and the deflection and rotation of the robot flange is measured in real time by the 6-DOF ray measurement device. This drives the robot controller in closed loop mode to adjust the current at each servomotor until the end effector goes back to its original position. This process is repeated for different loads, different directions and different configurations and eventually a 3-D look-up table is populated from the interpolated data between calibrated positions. During calibration the controller works in closed loop to determine the correct current to each servo motor and during operation, the controller works in open mode because the correct currents are already known.
Developmental status and risk assessment
Broadly speaking we do not have any unknowns here and all technologies required to make this concept happen are right out there; all we have to do is to apply those technologies in a lateral way. This is what this concept is all about. This automatically places this project a a low developmental risk.
Take for example the laser generator. The easiest way to develop multiple co-planar rays is to use a single laser source and split it into multiple beams using Diffractive Optical Elements (DOEs). These elements are extensively used in structured light applications and are mature technologies. Stoker-Yale has developed DOEs with 120 degrees fan angle and 5 degrees inter-beam angle. This is not a challenge. Custom-made DOEs can be made for less than 10,000 US dollars and 200-300 US dollars each afterwards. I expect to use 4 lasers in one plane and 4 in the other. Alternative, a single laser could be used. This source is initially split into a number of beams using optical fibres and subsequently split into multiple beams using DOEs.
The laser generator is a self-powered unit and does not have to acquire or transmit any data; It is just attached to the robot flange. Beam stability is extremely important. This is the very first item that has to be researched. The good thing is that in-line calibration takes only 3-4 minutes so lasers are periodically switched on so there is no time for temperature to built up. In addition, Quantum Dot lasers could be used because they are not affected by temperature. Finally, laser beams could be calibrated for different temperature ranges if required.
The base ring will be more expensive as about 32 imaging sensors will be required but the technologies are known.
Target selling price of the system less than 35,000 US dollars.
Resources requires as follows:- Myself as a full time developer and project co-ordinator, and another full-time person. Specialist expertise will be brought in as required.
Project planning, resource allocation and costs to be discussed with interested parties.
What are the key innovations?
We work with local references that make each robot an independent absolute manufacturing and measurement tool.
Instead of trying to network robots directly -an impossible task in my view- we network their local references and calibrate robots against the known and verified absolute positions of their local references. This solves the problem of robot networking.
The key concept is the new 6-DOF measurement device that uses laser beams as references against calibrated imaging sensors whose positions are known with respect to the absolute frame of reference. The technology already exists but is used in a lateral way. This sets out an avalanche of possibilities. Firstly, it makes possible to calibrate robots in line and deal with temperature related errors. Secondly, it solves the robot networking problem and by extension, the problem of large volume manufacturing and in-line inspection, since each robot is effectively an absolute manufacturing as well as measurement tool.
For stiffness related errors we use the 6-DOF laser measurement device to calibrate the robots in closed loop mode and in production, we operate the robots in open loop mode for ultra-quick response.
Finally, one of the most powerful points of this concept has to be its elegance and generality. The concept can be applied to any robot without modification and is neither robot or application specific