Carl M Stott
Carl M Stott
Kuka Quantec Nano Model
As part of my Algorithms for Sensor Based Robotics class, I developed a dynamics model of the Kuka Quantec Nano using screw theory in Matlab. I then programmed multiple inverse kinematics algorithms and examined how each algorithm affected the robot's ability to reach a desired pose. Ultimately, I crafted my own algorithm, enabling the robot to achieve a desired pose while adhering to joint limits. Note that the GIFs may appear a little jumpy; the best way to view them is by opening them as pop-up windows. the full Github repo for this project can be viewed
Newton-Raphson without damping
Newton-Raphson with damping factor of .1
Jacobian transpose with .001 damping, note how much smoother it is than Newton-Rhapson
Jacobian transpose trying to reach a singular position, operating near singularity is where Jac-transpose begins to fail
damped least squares method, trying to reach the same singular position as the Jacobian transpose algorithm. Note how while it moves less smoothly than Jacobian transpose, it operates much better near singularity. Ideally you can have your robot use Jac-transpose when not near a singularity, then swap to DLS when near a singularity.
my own optimization algorithm constrained to obey joint limits
Same algorithm without a joint limit constraint trying to reach the same point (clearly fails, showing that the algorithm constrained to obey joint limits works)