Learning control is a new approach to the probelm of skill refinement
for robotic manipulators. It is considered to be a mathematical model
of motor program learning for skilled motions in the central nervous s
ystem. This paper proposes a class of learning control algorithms for
improving operations of the robot arm under a geometrical end-point co
nstraint at the next trial on the basis of the previous operation data
. The command input torque is updated by a linear modification of pres
ent joint velocity errors deviated from the desired velocity trajector
y in addition to the previous input. It is shown that motion trajector
ies approach an epsilon-neighborhood of the desired one in the sense o
f squared integral norm provided the local feedback loop consists of b
oth position and velocity feedbacks plus a feedback term of the error
force vector between the reactive force and desired force on the end-p
oint constrained surface. It is explored that various passivity proper
ties of residual error dynamics of the manipulator play a crucial role
in the proof of uniform boundedness and convergence of position and v
elocity trajectories.