Active Interpolated Surfaces for Robotic Arm Control: The Potential for High Speed Implementation in FPGA
The main contribution of this paper is to introduce a novel control system, formulated to take advantage of recently developed FPGA hardware in application to robot arm control, to demonstrate the potential of FPGA for manufacturing processes. Robotic Arms are commonplace in manufacturing. Active interpolated surfaces offer easy implementation in FPGA hardware and thus have fast processing as well as speedy programming and modification times, allowing for wide areas of applicability. High speed implementation potentially offers flexibility and fast adaptation. This paper reports on an active interpolated surface model applied, as an illustrative example application, to the task of removal of flawed coin templates in a mint. The task requires fast location of flawed items in a wide field of variably positioned coins, and the speedy determination of position and movement requirements to enable the location and extraction of the flawed coins using PUMA robotic arm.
If I understand correctly, the FPGA controller is feeding simple (X,Y) coordinates to the robot. Presumably there's a sucker-up-down subroutine or some such to deal with the actual grabbing, and to ensure that the tool isn't driven through all the unflawed coins.
Does this mean that the system relies on the kinematic equations already in place on the robot? If so, does this unsupervised learning system run a chance of encountering singularities and crashing the arm? Or am I too used to older VAL controllers? ;)
Very interesting paper. It introduces a novel control system. However, it would have been better if the existing control systems are briefly reviewed and compared with the new system. This is necessary for justifying the selected approach and for validating the results.