© Copyright Notice: The contributing author(s) include the documents contained in these directories/web pages as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.
A new class of Intelligent Adaptive Control Designs for complex and multimodal dynamic systems is introduced.
Those systems under consideration include dynamic systems that show multi-modality, scheduled and unscheduled
‘Jumps’, as well as unmodeled dynamics, and are often under the challenge of unforeseen changes due to plant
dynamics and/or external influences. The focus of the dissertation is to develop Intelligent Supervisory Loops
augmenting a stable Direct Model Reference Adaptive Controller in order to control such systems. In view of
the above, four design formulations are developed which evolved from different methods of conceiving
Intelligent Supervisory Loops. Subsequently, those formulations are structured into intelligent control
algorithms and investigated with comprehensive simulation models of a Single Link Flexible Robotic
Manipulator as well as a 6-DOF F16 Fighter Aircraft. In order to facilitate these dynamic systems
variations, the control algorithms are applied on the developed models that are activated under
artificially created structural fluctuations.
Controlling dynamic systems, which are multimodal, shows sudden ‘Jumps’ and at the same time are susceptible
to un-modeled dynamics, are important as most of the complex dynamic systems are known for such behavior.
Some times the system mode switching and sudden ‘Jumps’ can be recognized a priori facilitating the designer
to develop a knowledge base offline there by using this database to generate the changing reference model.
In such a condition, a fuzzy reference model structure as developed in scheme one can be used effectively
once the knowledge base in the form of fuzzy inference engine can be constructed. On the other hand if the
designer have very little knowledge about these system characteristics then an online learning dynamic
reference model structure is necessary. The design of the next two higher-level intelligent control algorithms
stems from this established fact.
1. Fuzzy Multiple Reference Model Adaptive Controller(FMRMAC)
The first scheme is a novel fuzzy logic based multiple reference model adaptive control scheme for
multimodal and dynamic ‘Jump’ system. The proposed scheme consists of a fuzzy logic switching method
within the Model Reference Adaptive Control (MRAC) framework without using any explicit identifier.
The switching scheme is used for generating appropriate reference models on line so that effective
overall performance of the adaptive controller is achieved. The scheme is based on Takagi-Sugeno fuzzy
system and produces a ‘soft’ way of generating the reference model, combining a group of weighted
reference models effective at each modal operation. The scheme can therefore be implemented as a
Fuzzy Multiple Reference Model Adaptive Controller (FMRMAC). Following a rule base, the fuzzy switching
scheme effectively monitors changes in plant operating conditions and mode changes due to any sudden
‘Jumps’ in the plant. A fuzzy inference engine then fires appropriate rules, which gives fuzzified
output values. Defuzzification is then performed on line, monitoring the plant auxiliary states or
derived measurements.
The main contribution of such approach is that it can be performed online and is very well suitable
for applications that show sudden movements viz., ‘Jumps’ in the plant operating conditions.
Unlike static multiple model algorithms for switching (non-interacting individual model-based filters)
or switching dynamic algorithms (susceptible to numeric overflow), this scheme provides an interactive
multiple model environment with soft switching. The scheme is computationally feasible, effective and
efficient. Further this method can be enhanced by additional learning strategy to modify the rule base
depending on the expansion of the plant operating range. Moreover, due to its ability to functionally
represent the modes at each control interval from the combination of the modes obtained based on the
developed rules; the scheme is shown to yield a Fault Tolerant Controller.
2. Neural Network Parallel Adaptive Controller(NNPAC)
The second scheme is a novel Neural Network Parallel Adaptive Controller (NNPAC), which consists of an online
growing dynamic Radial Basis Function Neural Network (RBFNN) structure in parallel with a direct Model
Reference Adaptive Controller (MRAC), for complex and un-modeled dynamic system control. For the changing
modes of operation of the plant, a suitable single reference model is used, while the RBFNN controller
operating in parallel with the direct MRAC responds to those changing modes. Moreover, the parallel
RBFNN controller is capable of precisely tracking the system output to the desired command signal
generation. The update details of the RBFNN width, centers and weights are derived in order to ensure
the error reduction and for improved tracking accuracy.
The importance of the proposed scheme is its ability to perform effectively even when the plant mode swings
without using Multiple Model Concept or a Multiple Reference Model Adaptive Controller. Unlike the Model
reference adaptive control in which the plant follows the reference model output, the proposed scheme is
capable of preparing the plant output to follow the desired trajectory itself. Moreover, the RBFNN structure
avoids the ‘Dimensionality Problem’ inherent in such architecture and is dynamic and growing in nature.
Thus the RBFNN online scheme provides an additional control value dynamically inverting the plant
nonlinearities, such that the systems un-modeled dynamics are effectively controlled and correcting
the reference model output deviation from the command signal.
3. Neural Network Parallel Fuzzy Adaptive Controller(NNPFAC)
In the third scheme a Neural Network Parallel Fuzzy Adaptive Controller (NNPFAC) is presented.
This algorithm is suitable for the system which shows multi-modality and susceptible to un-modeled dynamics.
The growing RBFNN neural network augments the direct adaptive controller algorithm and generates a total
control, which is similar to that established in scheme two. However when the system shows sudden ‘Jumps’
and mode switching a sparse fuzzy inference engine generates a changing reference model structure there
by enabling a moving reference structure in the direction of the adaptive control alleviating the stress
on the controller. The algorithm shows interesting results when applied to both the dynamic nonlinear models
discussed before.
This scheme is important as it enjoys the dynamic learning capability of neural network especially in the
presence of system un-modeled dynamics. Moreover, the neural network algorithm prepares the system output
to precisely track the reference trajectory even in the presence of multiple reference models.
It also assuages the creation of dimensionally dense fuzzy system rule base as the mode swings are
learned effectively by the RBFNN and the need for the fuzzy multiple reference model generation is
only when the system undergoes mode switching or sudden ‘Jumps’. Thus, under the condition of the
system moving from one node to another, the capability of fuzzy switching reference model enables
to have a stable adaptive controller. Thus the overall scheme performs effectively and efficiently
when the systems mode swings which is known a priori.
4. Composite Parallel Multiple Reference Model Adaptive Controller(CPMRMAC)
In the presence of un-scheduled and unforeseen changes due to plant dynamics and/or external influences,
there is a need to generate reference model structures suitable for such changes and in real time of the
plant operation. Moreover, the reference model structure needs to evolve from monitoring the plant input
(auxiliary variables) and output. The fourth scheme is a Composite Parallel Multiple Reference Model
Adaptive Controller (CPMRMAC), which is suitable to control such systems. It consists of a Multi Layered
Perceptron (MLP) based offline neural network which approximates the reference model structural changes
while the system is under operation based on some previous knowledge. An online dynamic RBFNN produces
the correction to this change based on any unforeseen change in plant dynamics. This ensures a stable
adaptive control, which is suitable for mode variations, and unforeseen online plant dynamics. Due to
the presence of the online RBFNN, the offline MLP learning is not intense and suitable for a sparse
knowledge base. Therefore, these effects are addressed by a parallel RBFNN, which ensures that even
when there is un-modeled dynamics the overall closed loop system is stable and tracks the command
signal precisely.
Conclusion
The four design formulations developed in the research for conceiving
Intelligent Supervisory Loops into the MRAC framework show immense potential for the
control of complex systems as has been demonstrated by their application to two
challenging practical systems; the Pitch Rate dynamics of the fighter aircraft and the
Precise Tracking of a Flexible Link Robotic Manipulator. Even though the problem
attempted to demonstrate the abilities of the proposed schemes are specific in nature
they are the representative controls of a wide range of applications such as those
mentioned above. Looking forward, I would like to expand on both theoretical and
practical aspects of those Intelligent Controllers and to develop new schemes for
various inter and cross disciplinary applications, such as those related to Energy
systems, Bio systems, Micro controller and Real Time Embedded Systems applications
where I have been working on in the past.