**8. Conclusion**

112 Recurrent Neural Networks and Soft Computing

Fig. 16. The overall simulation of Humanoid Manipulator 27-DOF with Virtual Reality

movie of model in side the VR environment according to the path of inputs coordinate.

Fig. 17. a, b, c, d. The simulation results for posture of humanoid manipulator 27-DOF.



Based upon FRNN with VR model.

The results of the simulation can be display by two methods. The first by running the design as one iteration by input one set of position and orientation of the limbs where the results is the posture form of humanoid manipulator in VR same as results shown in Fig. 17.a,b,c,d. The second method is achieved by enter many sets of coordinates to design. The results were From the description of the all types of RNN, some of special modified in the main types to get special RNN structure related to special application such as the calculation of system parameters by identification for the states system, recognition of fault with experience system and From the description of main types of training RNN can be seen that the RNN is very flexible structure to apply many mathematical algorithms and implement to solving system problem. The implement of feedback technique with memory in RNN gives the ability to dealing with many problems that needs high calculation with iteration to get the solution.

The solution of IKP is achieved by different method but the powerful method with practice application is the analytical solution. The calculation of posture for human body or manipulator which has the similar kinematic structure with human can be achieved by analytical solution, but this solution is very difficult because of high mathematic and very length calculation. This difficulty is increasing when we need calculated the posture with time to get the movie of manipulator. This problem is solving in this chapter by using FRNN with back propagation training algorithm.

The results of calculation are achieved for each limb separately at beginning such as for arm, lag and neck. After checking the accuracy of results by applying the forward kinematic to get the same coordinate of end- effecter for each part. The overall calculation of IKP is achieved to get the posture of manipulator. The data base for IKP is used to identify the FRNN for all point of envelop for movement where the internal memory and feedback in FRNN assistance this structure to overcame the problems of high calculation in the IKP solution.

The link between Matlab / Simulink and VR environment and the suitable of Matlab to execute any algorithm with high calculation are assistance to implement the posture of manipulator with high accuracy for two cases of running (movies and once posture). *The future* works for this design are:

