On the tuning of the lagrangian parameters in a uniform structure in adaptative robot control

Rudas, Imre J.; Tenreiro Machado, J. A.; Kaynak, Okyay M.

http://hdl.handle.net/10400.22/13562

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
COM_MachadoTenreiro66.pdf		333.22 KB	Adobe PDF	Download

Send Feedback

Authors

Rudas, Imre J.

Tenreiro Machado, J. A.

Kaynak, Okyay M.

Abstract(s)

In this paper the application of "uniform structures" formerly introduced for the Hamiltonian description of robtos is investigated in the case of a SCARA within the frames of the Lagrangian model. This model does not suffer from measurability problems of the Hamiltonian one. Via simulation it was found that this simple method can improve the control of an imperfectly modeled system under unmodeled environmental interaction. It uses the Lie parameters of the Orthogonal Group and the "Sliding Simplex Algorithm" for the on-line tuning the parameters in the uniform structures. This method is very similar to the learning process of the artificial neural Networks. To evade the problem of local optima tuning starts from the estimtaed vicinity of the Global Optimum at the begining. It is shown that a fast enough tuning can "stick" in this optimum and it propagates together with it in time as the local dynamics of the system is changes in time.