No description available.

Robot control systems can be considered complex systems, the design of a controller involving the determination of the dynamic model for the system. This in itself can be a complicated task due to non-linearities, multiple axis (degrees of freedom) control and the constantly changing working environment. Problems arise when the theoretical model produced for such a system is not accurate. When developing a controller using conventional techniques a design scheme has to be produced, usually based on a model of the system. In addition kinematics equations must be derived to take into account the physical boundaries of the system. The work outlined in this thesis utilises fuzzy logic control to address these control issues. Fuzzy logic provides functional capability without the use of a system model and has characteristics suitable for capturing the approximate, vaiying values found in real world systems. Initial development of a single axis fuzzy logic control system was implemented on a Dainichi industrial five-axis robot, replacing the existing control and hardware systems with a new developmental system. The concept of fuzzy logic and its application to control highlights the potential advantages that fuzzy logic control (PLC) can provide when compared to the more conventional control methodologies. Additional new control hardware has been interfaced to an existing robot manipulator, making it possible to compare PLC and PIDVF (Proportional Integral Derivative Velocity FeedforwardlFeedback) controllers for single axis development. Average response time and overshoot for a given set point were compared for each system. The results proved that, using a basic PLC minimal overshoot and fast rise times could be achieved in comparison to the commercial PIDVF system. Further research concentrated on the development of the control software to provide multiple axis control for an industrial robot using a continuous path algorithm. The more from single axis to multiple axis control provided a much more complex control problem. A novel and innovative process for the fuzzy controller was implemented with up to three axes reaching the target point simultaneously. Control of the industrial robot was investigated using methods that were more suited to real time controL The most significant change was a reduction in the number of fuzzy rules when compared to single axis control. During robot control no adaptation of the rule base or membership functions was carried Out Ofl line; only system gain was modified in relation to link speed and joint error within predetermined design parameters. The fuzzy control system had to manage the effects of frictional and gravitational forces whilst compensating for the varying inertia components when each linkage is moving. Testing based on ISO 9283 for path accuracy and repeatability verified that real time control of three axes was achievable with values of 938tm and 864tm recorded for accuracy and repeatability respectively. The development of novel industrial robot real time multi-axis fuzzy controller has combined new control hardware with an efficient fuzzy engine addressing inverse kinematics, scaling and dynamic forces in order to provide a viable robot control system.

This programme of research was supported by NEI Parsons Ltd. who sought a robotic means of polishing mechanical components. A study of the problems associated with robot controlled surface processing is presented. From this evolved an approach consistent with the formalisation of the demands of workpiece manipulation which included the adoption of the Hybrid robot control scheme capable of simultaneous force and position control. A unique 3 axis planar experimental manipulator was designed which utilized combined parallel and serial drives. A force sensing wrist was used to measure contact force. A variant of the Hybrid control 'scheme was successfully implemented on a twin computer control system. A number of manipulator control programs are presented. The force control aspect is shown both experimentally and analytically to present control problems and the research has concentrated on this aspect. A general analysis of the dynamics of force control is given which shows force response to be dependent on a number' of important parameters including force sensor, environment and manipulator dynamics. The need for a robust or adaptable force controller is discussed. A series of force controlled manipulator experiments is described and the results discussed in the context of general analyses and specific single degree of freedom simulations. Improvements to manipulator force control are suggested and some were implemented. These are discussed together with their immediate application to the improvement of robot controlled surface processing. This work also lays important foundations for long term related research. In particular the new techniques for actively controlled assembly and force control under 'fast' operation.

Robot control systems evolved with genetic algorithms traditionally take the form of floating-point neural network models. This thesis proposes that digital control systems, such as quantised neural networks and logical networks, may also be used for the task of robot control. The inspiration for this is the observation that the dynamics of discrete networks may contain cyclic attractors which generate rhythmic behaviour, and that rhythmic behaviour underlies the central pattern generators which drive lowlevel motor activity in the biological world. To investigate this a series of experiments were carried out in a simulated physically realistic 3D world. The performance of evolved controllers was evaluated on two well known control tasks—pole balancing, and locomotion of evolved morphologies. The performance of evolved digital controllers was compared to evolved floating-point neural networks. The results show that the digital implementations are competitive with floating-point designs on both of the benchmark problems. In addition, the first reported evolution from scratch of a biped walker is presented, demonstrating that when all parameters are left open to evolutionary optimisation complex behaviour can result from simple components.