Through the investigation of the pedagogical robotic products currently on the market it seems that there are no mainstream systems that deal with the specific reconfigurable nature described in this thesis. The notion that this may be known conclusively however is not one that may survive further investigation. As such, the Massey R&D and IP teams were employed to further examine novelty of the proposed concept, and at the time this thesis was written, the patent search confirmed the originality of the Shell and Core robotic manipulator concept.
The conducted research achieved the set goals as defined in Section 1 by developing a Shell-Core reconfigurable robotic manipulator concept and producing a physical prototype that includes the following features:
x Compact joint control system. x Quick-change modules.
x Robotic manipulator configuration auto-identification. x Plug-and-Play.
x Position tracking. x Neat hidden wiring.
Testing of the prototype proved the proposed concept and demonstrated that the system could change from one robot arm configuration into another in less than three minutes to perform new tasks.
While this research produced promising outcomes, there are several aspects that may benefit from improvement and further research, such as the implementation of artificial intelligence for job identification, part holding and path planning. Based on the development of the prototype, the considered control theories relating to the implementation of the proposed system provided a significant insight into the range and type of control methodologies appropriate to this area of discourse. Although a smart PI controller with some fuzzy elements was used as the main type of controller for the rotational actuator, it is not unreasonable to assume that a larger dependence may be put on fuzzy logic when implementing control for an assembly where response is affected by gravity in a non-uniform fashion. Nonlinearities are introduced into the response of the system as soon as a link is added vertically (fig. 9 - 1).
From briefly testing the smart PI controller in all of the above configuration (fig. 9 - 1), it is clear that the controller in its current state is not capable of automatically handling the non-linear response of the system to a precise degree, often overshooting the desired set- point and settling to a position within 30 encoder counts for HP control and 70 for MP control.
This was mitigated by manually selecting the correct velocity profile. Initially, this was a consideration and it was planned to implement a sliding fuzzy logic element to the controller that would automatically determine both proportional and integral gains depending on distance from the vertical (fig. 9 - 1), but due to time constraints, this can only be implemented in the next stage of this research project.
A drawback in terms of control for the reconfigurable robotic manipulator prototype was the backlash introduced by the motor driven gear and pinion mechanism (fig. 9 - 2). Although final steady-state accuracy in the SCARA configuration was not affected, steady- state rigidity was somewhat compromised.
Figure 9 - 2. Gear backlash diagram.
Play from the gear alone was found to be approximately 10 encoder steps (fig. 8 - 20). Although this backlash could be prompted by the smallest of disturbances during steady- state, it did not seem to be an issue during normal operation, nor did an added load affect control accuracy. If the actuator was disturbed during position tracking, final steady-state position was not affected. It seems that a rotational actuator’s lack of rigidity is fairly unimportant when the current manipulator operates in two planes only, such as the SCARA configuration, and is not disturbed when stationary. Gear backlash is almost removed completely when dealing with a configuration of the form shown in Figure 9 - 1, as there is constant force acting on the actuator. However, this is not the case when the arm is perfectly vertical. Of course, these singularity positions can be avoided through control methodologies.
The resulting system platform concept was found to produce very encouraging results, capable of achieving all mainstream industrial robotic manipulator configurations as defined in Section 3.1, given enough building elements. The end prototype is very promising in terms of proof-of-concept. Communication and power infrastructure is very flexible, providing a base system prototype capable of accommodating a range of future effectors, add-ons, supplementary software and advanced GUI control without a significant amount of additional work.
This basis already allows for ease of control through a GUI capable, PC based program such as Matlab through serial commands, setting the precedent for numerous advanced
control concepts of the manipulator for various applications. Without further improvement or development to the prototype system, it is possible to implement proprietary software that may calculate the reverse kinematics of the arm, based on information known about specific dimensions of the configuration. It is also possible to implement trajectory control of the manipulator. The notion of ‘teaching’ the manipulator arm what to do is already within the capability of the prototype. If all non-self-locking actuators are used, the user may move the arm to a desired location. Software may then request the current position of each arm and so a sequence of locations and tasks may be established.
The prototype’s repeatability was tested to be within ± 0.5 mm for the SCARA configuration. The intended purpose of this prototype did not incorporate use within a high accuracy industrial context; therefore this manipulator accuracy is acceptable. Taking this into consideration, it may be said that the prototype provided proof that the concept is viable and realistically achievable. Furthermore, accuracy, rigidity and manipulator strength may all be improved given a larger budget.