Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Neuro-fuzzy control (Adaptive Neural Fuzzy Inference System. ANFIS) onrobotic arm of 3 to 5 degrees of freedom. Modelling, kinematics, dynamics,simulation, evaluation

Gkionis-Konstantatos Odysseas

Full record

---

Summary

The industrial robotic arms are mainly installation and operation devices. The basic issue in controlling is the calculation of the robotic arm competence degree and more specifically the optimum design of movement control so that it follows a preferred orbit within the kinematics constraints.The conventional control methods are highly dependable on the precision mathematical modelling, analysis and synthesis. Approaches of this kind are suitable for the control of robotic systems which operate in structured environments. Nevertheless, for the functions in unstructured environments it is demanded to execute more complex duties without a suitable analytical framework. The most difficult problem in this domain regarding the unstructured surroundings, is that there are always uncertainties. These uncertainties are mainly caused by the sensors’ inaccuracy and the unpredictability of the environmental characteristics and dynamics. On the other hand, the advent of technical fuzzy aggregations provides a powerful tool for the resolution of demanding reality problems within uncertain and unforeseeable environments. The fuzzy controller can characterize better behavior compared to the classic linear controller PID because of its non-linear features. In principle, a rigid robotic arm of N freedom degrees is characterized by N non-linear dynamic, conjugated differential equations. The control problem of robot operators continues to provide many practical and theoretical challenges due to the complexity of robot dynamics and the advantages dictated through achievement of high precision orbit in the cases of high-speed mobility and highly transformed loads.In the context of this thesis, it is presented a new approach encountering a robotic arm with 3 to 6 freedom degrees adopting the control method ANFIS to safeguard the strategy of robot control position. Initially, it the complete kinematic, differential dynamic of the robotic arm is examined. The dynamic is exceptionally non-linear with strong connections existing between the joints and they are frequently placed in both structured and unstructured uncertainties. Fuzzy Logic Controller can very well describe the preferred behavior of our system, with simple relationships, as the designer gets rules manually through trial and error. On the other hand, the Neural Networks execute operational approaches of a system, but they can neither interpret the resulting solution nor control if their solution is reasonable. These two approaches are complementary. Combining them, the Neuronic Networks will allow the learning ability while Fuzzy-Logic will bring the representation of knowledge (Neuro-Fuzzy). The arithmetic simulation adopting the dynamic model captures the effectiveness of the approach in relation to the issues arising from orbit monitoring. Furthermore, the robotic arm will be designed in a dimensionally and kinematically accurate 3-D model with the use of CAD software program. Simulation, based on this model, will take place and will function not as a plain representation of the movement mechanism but it will also calculate the torques developed in the joints taking into account their given angular displacement. The resulting outcome will present the effectiveness and robustness of the proposed control and evaluation of the compatibility among the scores of calculation methods, experiments and simulations will follow.

Available Files

Services