Sammanfattning
This thesis deals with flexibility in the design of real-time control systems. By dynamic resource scheduling it is possible to achieve on-line adaptability and increased control performance under resource constraints. The approach requires simulation tools for control and real-time systems co-design.
One approach to achieve flexibility in the run-time scheduling of control tasks is feedback scheduling, where resources are scheduled dynamically based on measurements of actual timing variations and control performance. An overview of feedback scheduling techniques for control systems is presented.
A flexible strategy for implementation of model predictive control (MPC) is described. In MPC, the control signal in each sample is obtained by the solution of a constrained quadratic optimization problem. A termination criterion is derived that, unlike traditional MPC, takes the effects of computational delay into account in the optimization. A scheduling scheme is also described, where the MPC cost functions being minimized are used as dynamic task priorities for a set of MPC tasks.
The MATLAB/Simulink-based simulator TrueTime is presented. TrueTime is a co-design tool that facilitates simulation of distributed real-time control systems, where the execution of controller tasks in a real-time kernel is simulated in parallel with network transmissions and the continuous-time plant dynamics. Using TrueTime it is possible to study the effects of CPU and network scheduling on control performance and to experiment with flexible scheduling techniques and compensation schemes. A general overview of the simulator is given and the event-based kernel implementation is described.
TrueTime is used in two simulation case studies. The first emulates TCP on top of standard Ethernet to simulate networked control of a robot system. The second case study uses TrueTime to simulate a web server application. A feedback scheduling strategy for QoS control in the web server is described.
One approach to achieve flexibility in the run-time scheduling of control tasks is feedback scheduling, where resources are scheduled dynamically based on measurements of actual timing variations and control performance. An overview of feedback scheduling techniques for control systems is presented.
A flexible strategy for implementation of model predictive control (MPC) is described. In MPC, the control signal in each sample is obtained by the solution of a constrained quadratic optimization problem. A termination criterion is derived that, unlike traditional MPC, takes the effects of computational delay into account in the optimization. A scheduling scheme is also described, where the MPC cost functions being minimized are used as dynamic task priorities for a set of MPC tasks.
The MATLAB/Simulink-based simulator TrueTime is presented. TrueTime is a co-design tool that facilitates simulation of distributed real-time control systems, where the execution of controller tasks in a real-time kernel is simulated in parallel with network transmissions and the continuous-time plant dynamics. Using TrueTime it is possible to study the effects of CPU and network scheduling on control performance and to experiment with flexible scheduling techniques and compensation schemes. A general overview of the simulator is given and the event-based kernel implementation is described.
TrueTime is used in two simulation case studies. The first emulates TCP on top of standard Ethernet to simulate networked control of a robot system. The second case study uses TrueTime to simulate a web server application. A feedback scheduling strategy for QoS control in the web server is described.
Originalspråk | engelska |
---|---|
Kvalifikation | Licentiat |
Tilldelande institution |
|
Handledare |
|
Tilldelningsdatum | 2003 dec. 12 |
Förlag | |
Status | Published - 2003 |
Ämnesklassifikation (UKÄ)
- Reglerteknik