For robotic systems tracking a given geometric path, this paper addresses t
he problem of satisfying input and state constraints, According to a predic
tion of the evolution of the robot from the current state, a discrete-time
device called path governor (PG) generates on line a suitable time-paramete
rization of the path to be tracked, by solving at fixed intervals a constra
ined scalar look-ahead optimization problem. Higher level switching command
s are also taken into account. by simply associating a different optimizati
on criterion to each mode of operation. Experimental results are reported f
or a three-degree-of-freedom PUMA 560 manipulator subject to absolute posit
ion error, Cartesian velocity, and motor voltage constraints.