Optimal and Robust Control for Synchronization of Networked Multi-Agent Systems

摘要

Synchronized multi-agent systems can perform complex tasks often with precision more than what humans can achieve. In teleoperation, brain-computer interface (BCI) systems, preciseness in motion of the various coordinated axes of the robot is key for successful execution of tasks. However, uncertainties, disturbances, unknown environment and nonlinearities such as communication delay remain major issues and the solution to these problems is open. For human-in-loop systems like telesurgery, the additional delay from human reaction time apart from communication signal delay needs to be considered for effective control design of such time delay systems. Additionally, teleoperated robots are mostly battery-powered and, therefore, energy consumption rate in relation to the work done is critical. It is, therefore, imperative to design control systems considering some performance index.;In this research, a new method is presented to synchronize n-agents optimally considering a performance index. This provides an analytical way to optimally address the multi-agent synchronization problem. In other words, the method aims not just looking at stabilizing our system but considering cost to achieve that goal. Another issue is, to design control for our systems, the system is typically modeled from first principles of physics, using appropriate assumptions and simplifications. The final model, therefore, may not adequately capture the true physical system. There are also issues of environmental changes, changes in system dynamics, structural damages to the system. These are usually not considered in general control systems design. Consequently, it is important to design controls that can address these uncertain changes to ensure stable, secure and smooth operation under all conditions.;Sensors read actuation information and send feedback signals to controllers for correction. Sensor reaction time and network communication introduce communication delay in the controlled system. Communication delay has a detrimental effect on the performance of controlled systems degrading performance, and even resulting in instabilities. Time delay term results in an infinite number of roots of the characteristic equation making it difficult to check stability and design stabilizing controller. The research presents a method of designing controllers considering communication delay.;First, an optimal synchronization controller is designed using Linear Quadratic Regulation. Modeling and simulations are done in MATLAB/Simulink and experiments are performed to validate the results. The typical optimal controller successfully stabilizes but shows large tracking error, a PID-controller is therefore integrated in the optimal synchronization controller to reduce the error significantly.;Secondly, synchronized model reference adaptive controller is also designed and simulated through MATLAB/Simulink and validated with an experiment. External load disturbance is introduced through eccentric loading of one agent and the adaptive performance is evaluated.;Thirdly, effects of communication delay on Internet-based teleoperation of carts using UDP protocol is investigated with two quanser cart systems and a suitable controller designed with MATLAB/Simulink is used to simulate the controller and validated with an experiment through the quanser cart systems.;Lastly, a control design investigating the effects of communication delay on systems with brain-computer-interface (BCI) or human-in-loop is conducted using the WIDOWX arm robot from Trossenrobotics. Robot Operating Systems (ROS), an open source software working on Ubuntu, a linux-based operating system is used for this research.;Precision in multi-agent synchronization systems ensures effective synchronization, transient oscillations which is undesired affects the degree of preciseness. To reduce the magnitude of the oscillation and ensure quick decay to steady-state with adaptive capabilities that ensures continuous desired output characteristics, a new adaptive synchronization technic with a synchronization controller is proposed. The approach ensures approximately 99.3% reduction in synchronization error compared to synchronization without the controller. It also ensures 97--99.3% improvement in disturbance rejection properties. The continuous pole place is used to design a controller with the ability to stabilize the agents under the effects of time delay. Experiment show stabilized output with the controller and an instable output without the controller. The results from these are geared towards effective coordination of brain-controlled limbs and teleoperation. A typical example is the application in motor rehabilitation after stroke by decoding movement attempts. The oscillatory brain signals move supporting robotic/orthotic devices or virtual hands to close the feedback loop. Experiment with the robot arm shows the effect of human reaction delay on the control system.