Use this URL to cite or link to this record in EThOS:
Title: Knowledge-based control of robotic manipulators
Author: de Silva, C. W.
Awarding Body: University of Cambridge
Current Institution: University of Cambridge
Date of Award: 1988
Availability of Full Text:
Full text unavailable from EThOS.
Please contact the current institution’s library for further details.
The work described here considers the integration of knowledge-based soft control with hard control algorithms. Specifically, the development of a knowledge-based controller for robotic manipulators is addressed. Servo control alone is known to be inadequate for high-speed high-precision robots. Furthermore, knowledge-based control such as fuzzy control, when directly included in the servo loop, has produced unsatisfactory performance in research robots. These considerations along with the fact that human experts can very effectively perform tuning functions in process controllers, form the basis for the control structure proposed in this work. The proposed control structure consists of a three-level hierarchy. The lowest level has a set of PID controllers closed around a high-speed decoupling and linearising controller. A recursive algorithm has been developed for implementation of this nonlinear feedback controller. The second level contains a set of knowledge-based controllers known as servo experts. There is one servo expert for each degree of freedom of the robot. A servo expert is a knowledge system of the forward production type. It monitors the corresponding joint response and makes inferences on the basis of a set of performance specifications. These inferences are supplied to the fuzzy controller at Level 3. The knowledge base of the fuzzy controller consists of expert knowledge in the form of linguistic rules for servo tuning. These rules are expressed as fuzzy decision tables by using membership functions of the condition and action variables, fuzzy logic, and the compositional rule of inference. This knowledge, along with external sensory data and other available information, is used to arrive at tuning decisions for the PID controllers. Control specifications, parameters of the decoupling controller, and the rule bases themselves can be modified as well, using this type of knowledge, at the fuzzy control level. Separation of the knowledge-based control into two levels, with the lower level functioning as an intelligent preprocessor and the upper level containing a fuzzy knowledge base that is representative of expert knowledge in servo tuning, is an important characteristic of the proposed control structure. The background work of the present research includes a thorough literature review on hard algorithmic control and knowledge-based soft control of robotic manipulators. Theory, concepts, and procedures for developing each level in the proposed hierarchical control structure, have been established. As an application of the proposed control structure, a knowledge-based control system has been developed for a two-degree-of-freedom robot. The system contains a servo expert for each joint of the robot, developed using a commercially available Al toolkit (MUSE). The unstructured code in the servo experts was written in PopTalk language and the structured code was developed using the editor-tool facility of MUSE. The fuzzy controller was developed off line and it has been implemented as a set of decision tables. A robot simulator was developed as a separate UNIX process written in C language. The simulator has been interfaced with the servo experts and with the fuzzy controller using the UNIX Socket facility, and the Channel objects and Data Stream objects that are available with the Al toolkit. Data channel programs were written in PopTalk. Performance of the overall system was evaluated using simulation experiments. Simulation experiments which were carried out included single-joint tests with step, ramp, and sine inputs. Results have demonstrated the superiority of the proposed control approach in comparison with conventional control without knowledge-based tuning. Simulations of a series of seam tracking tasks were carried out with the complete robot. Initial position errors were introduced and acceleration disturbances were injected during operation, fast negotiation of a corner, and operation under an imperfect nonlinear feedback controller. In each simulation, superior results were achieved with the proposed controller, in comparison to a conventional controller. Even though the performance generally improved with the bandwidth of the knowledge-based controller, a bandwidth of at least an order of magnitude smaller than the servo bandwidth consistently provided good performance. The controller was also found to be robust in the sense of having relative insensitivity to initial values, and other parameter values of the controller.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available