arXiv subjects

Davood Gharavian

Publications and source records attributed to Davood Gharavian.

7 recordsLinked to original sources

arXiv2020

Hardware in the loop simulation and hierarchical predictive control of a TWPTR

In this paper, we use a nonlinear hierarchical model predictive control (MPC) to stabilize the Segway robot. We also use hardware in the loop (HIL) simulation in order to model the delay response of the wheels' motor and verify the control algorithm. In Two-Wheeled Personal Transportation Robots (TWPTR), changing the center of mass location and value, the nonlinearity of the equations, and the dynamics of the system are topics complicating the control problem. A nonlinear MPC predicts the dynamics of the system and solves the control problem efficiently, but requires the exact information of the system models. Since model uncertainties are unavoidable, the time-delay control (TDC) method is used to cancel the unknown dynamics and unexpected disturbances. When the TDC method is applied, the results show that the maximum required torque for engines is reduced by 7%. And the maximum displacement of the robot has dropped by 44% around the balance axis. In other words, robot stability has increased by 78%. Due to the cost of implementing control in practice, this research runs the HIL simulation for the first time. The use of this simulation helps in implementing the control algorithms without approximation, and also the system response can be discussed in a more realistic way.

eess.SY↗

arXiv2020

Optimal control of signalized intersection using hierarchical fuzzy-real control

This paper presents a method based on precise modeling of traffic flow using a fuzzy-real algorithm for optimal control of a signalized intersection. By improving social indicators such as security, well-fare and economy, intercity transportation has increased and traffic congestion developed. Today, the plight of traffic congestion is the cause of waste of time, capital, and health risks. Although the construction of new streets, expressways, and highways can improve this situation, when inadequate space and limited budget is available, it is not practical. Optimization of signalized intersections for optimal use of the network capacity is proposed as a useful solution to resolve an important part of the problems developed at intersections. In this paper, with the aim of assessing the algorithms proposed for controlling signalized intersections and given the real potentials of common sensors present in traffic control systems, a suitable simulation medium called Signalized Intersection Fuzzy-Real Control (SIFRC) has been designed and implemented. Through the possibility made, the Abshar intersection in Isfahan city was modeled and phased through six methods: fixed time, pre-time control, segmental pre-time, fuzzy, real-time, and fuzzy-real. The results indicated better performance of the fuzzy-real method compared to the fixed time method up to 50%, and up to 10% compared to other methods. In addition, the speed of achieving the optimal response in the fuzzy-real method has been 11 times as large as that of the real-time method, while the delay time developed by the intersection in this method did not increase compared to the real-time method.

eess.SY↗

arXiv2017

On Coparanormality in Distributed Supervisory Control of Discrete-Event Systems

Decomposition and localization of a supervisor both are reduction methods in distributed supervisory control of discrete-event systems.Decomposition is employed to reduce the number of events and localization is used to reduce the number of states of local controllers.In decomposition of a supervisor both observation and control scopes are restricted, whereas in localization only control authority is restricted to the corresponding local controller. In this paper, we propose a decomposition method by defining coparanormality property, and by using relative observability property of a monolithic supervisor. Coparanormality is a coobservation property defined based on paranormality property for a set of natural projections.It is shown that each supervisor can be coparanormal, provided a set of appropriate natural projections exist. Moreover, it is proved that relative observability is a sufficient condition for decomposition of a supervisor. Furthermore, the supervisor localization procedure is generalized to find a set of local controllers for any partition ofthe controllable events set. The implementation of such local controllers may become easier in industrial systems.

eess.SY↗

arXiv2017

Localization of DES Supervisory Control with Event Reduction

Supervisor localization procedure can be employed to construct local controllers corresponding to component agents in discrete-event systems. The proposed method in [11] is based on state reduction of a monolithic supervisor with respect to each set of controllable events corresponding to each component agent. A supervisor is localizable if state cardinality can be reduced from the reduced supervisor to each local controller. Although event reduction is an important property, the original supervisor localization procedure did not guarantee event reduction in each local controller comparing to the reduced supervisor. In this paper, we propose a method to localize a supervisor with event reduction in each local controller comparing to the reduced supervisor. State reduction facilitates the implementation of local controllers on industrial systems, whereas event reduction reduces communication traffic between each pair of local controllers.

eess.SY↗

arXiv2017

On Preserving Observation Properties of the Reduced Supervisor in Discrete-Event Systems

Supervisor reduction procedure can be used to construct the reduced supervisor with a reduced number of states in discrete-event systems. The main concepts which are used in this procedure are control consistency of states, control cover, induced supervisor, and normality of the reduced supervisor w.r.t. the original supervisor. In this paper, it is proved that the reduced supervisor, constructed by the proposed method in [9], preserves the observation properties, i.e. normality and relative observability, by self looping corresponding unobservable events at some states of the reduced supervisor. This property can be applied to find a natural projection, under which the supervisor is relative observable.

eess.SY↗

arXiv2017

Optimal Non-blocking Decentralized Supervisory Control Using G-Control Consistency

Supervisory control synthesis encounters with computational complexity. This can be reduced by decentralized supervisory control approach. In this paper, we define intrinsic control consistency for a pair of states of the plant. G-control consistency (GCC) is another concept which is defined for a natural projection w.r.t. the plant. We prove that, if a natural projection is output control consistent for the closed language of the plant, and is a natural observer for the marked language of the plant, then it is G-control consistent. Namely, we relax the conditions for synthesis the optimal non-blocking decentralized supervisory control by substituting GCC property for L-OCC and Lm-observer properties of a natural projection. We propose a method to synthesize the optimal non-blocking decentralized supervisory control based on GCC property for a natural projection. In fact, we change the approach from language-based properties of a natural projection to DES-based property by defining GCC property.

eess.SY↗

arXiv2016

On Event Reduction in Localization of DES Supervisory Control

Supervisor localization procedure can be used to construct local controllers corresponding to each component agent in discrete-event systems. This procedure is based on state reduction of a monolithic supervisor with respect to each set of controllable events corresponding to each component agent. State reduction from the reduced supervisor to each local controller is an important criterion. In this paper, we deal with event reduction. It is proved that the number of events in each local controller is less than the event cardinality in the reduced supervisor, provided that each local controller has less number of states comparing to the reduced supervisor. It shows that each local controller needs less number of events to make consistent decisions. Having this property, we can evaluate the localizability of a supervisor by event reduction criteria instead of state reduction criteria. State reduction facilitates the implementation of local controllers on industrial systems. Whereas, event reduction reduces communication traffic between each pair of local controllers.

eess.SY↗