A numerical method of predetermined optimal resolution for a redundant manipulator

For the application of robot manipulators to complex tasks,it is often necessary to control not only the position of a manipulator butalso the force exerted by the end-effector on an object. For this purpose,Raibert and Craig proposed the hybrid positiodforce control method.Extending their method, it is proposed that the dynamic hybrid controlmethod, which takes into consideration the manipulator dynamics and theconstraints on the end-effector specified by the given task. One difficultyin implementing the method is that we usually do not have preciseinformation on the size and position of the object with which the end-effector contacts. To cope with this difficulty, a problem of dynamic hybridcontrol with unknown constraint is studied. After a brief introduction ofthe dynamic hybrid control approach, an on-line estimation algorithm isdeveloped that estimates the local shape of the constraint surface by usingmeasured data on the position and force of the end-effector. It is thenshown by experiments, using a SCAM-type robot, that the combinationof this algorithm with the dynamic hybrid control method works fairlywell. This approach decreases the burden on the operator of giving precisedata on the constraint and makes the dynamic hybrid control approachmore practical.