You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Learned trajectory-level inverse kinematics for the PUMA 560: set-prediction network, learned refiner and Viterbi DP, benchmarked against closed-form IK
Python robot kinematics toolkit for a planar 3R arm, a SCARA and a PUMA 560: standard DH forward kinematics, closed-form and damped-least-squares IK, Jacobians and singularities, Monte Carlo workspaces, PTP/LIN trajectories and certified capsule collision checks.