Motion design for high speed machines
The dynamic performance of a programmable manipulator depends on both the motion profile to be followed and the feedback control method used. To improve this performance the manipulator trajectory requires planning at an advanced level and an efficient control method has to be used. The purpose of this study is to investigate high-level trajectory planning and trajectory tracing problems. It is shown that conventional trajectory planning methods where the motion curves are generated using standard mathematical functions are ineffective for general application especially when velocity and acceleration conditions are included. Polynomial functions are shown to be the most versatile for these applications but these can give curves with unexpected oscillations, commonly called meandering. In this study, a new method using polynomials is developed to overcome this disadvantage. A general motion design computer program (MOTDES) is developed which enables the user to produce motion curves for general body motion in a plane. The program is fully interactive and operates within a graphics environment. A planar manipulator is designed and 'constructed to investigate the practical problems of trajectory control particularly when operating at high speeds. Different trajectories are planned using MOTDES and implemented to the manipulator. The precise tracing of a trajectory requires the use of advanced control methods such as adaptive control or learning. In learning control, the inputs of the current cycle are calculated using the experience of the previous circle. The main advantage of learning control over adaptive control is its simplicity. It can be applied more easily in real time for high-speed systems. However, learning algorithms may cause saturation of the driving servo motors after a few learning cycles due to discontinuities being introduced into the command curve. To prevent this saturation problem a new approach involving the filtering of the input command is developed and tested.