Fault tolerant control systems

From CajunBot Wiki
Jump to: navigation, search

End of Semester report on Fault Tolerant Control system of CAJUNBOT written by Darline Devariste for Dr Afef Fekih in EECE 599 class.

Fig 1. Architecture of FTCS


[edit] Introduction

Research on fault tolerant control (FTC) has been increasing in the last few years because FTC system has the ability to increase complex systems reliability and performance requirement in the events of faults. The design of a FTC system requires knowledge of advanced control mechanism. Systems mostly are very complicated. Designing a FTC system could also be very challenging. Different types of faults such as actuators, sensors, and system faults can occur. Each type of fault requires different approach to work with.

A fault tolerant control system must be able to perform: fault detection, fault isolation, and fault diagnosis. FTC should also have the ability to detect faults and provide correction. Fault tolerant control system results on two approaches: active and passive. The active approach relies on fault detection and isolation (FDI) scheme to detect the occurrence of faults in the system and to identify the source and severity of the faults [1]. Secondly, in passive FTC, potential component faults are known a priori and are all taken into consideration in the control system design stage[2].

The control law of the passive FTC is designed offline; it can be used instantly to perform fault detection, diagnosis and isolation. However, the passive approach cannot be used as the basic controller; it cannot guarantee the optimal performance of the nominal system. Furthermore, the controller might not fit the system with online faults. [2][1][3].The control law of the active FTC system is designed to operate when the system is running. This approach deals better with online faults. The challenging parts of an active FTC is designing a system that is robust enough to maintain system stability when the system is in reconfiguration mode. The FDD unit should be also robust to minimize error detections.[2][1][4][5].

Fault tolerant control has been used largely to ensure aircraft safety, automobiles, and several computing systems. Therefore it plays a very important role in the areas of process control, communications, transportations, and e-commerce, and space. The design of Aircraft requires systems that are extremely reliable. For that reason, Fault tolerant control is mandatory to the safety of passengers. So many things can go wrong while operating an aircraft or a vehicle. [6]

Cajunbot is equipped with multiple computers that can fail anytime. Therefore fault tolerance control is mandatory to the safety, reliability, and performance of the vehicle. An autonomous vehicle like the cajunbot must perform efficiently under system uncertainty within the plant and the surrounded environment for extended time. It should be able to compensate for plant, actuator, and sensor failures without any external intervention. An autonomous robust FTCS is developed in [2] that can be suitable for cajunbot. This controller combines the advantages of both passive and active approach of FTCS. The flowchart below explains the objective of the proposed controller in [2].

Fig 2: Scheme of the autonomous robust reliable controller.

[edit] Failure Analysis of Cajunbot

First let’s analyze all possible faults or failure that the vehicle might encounter. Autonomous vehicles need very accurate control. In order to achieve reliable and effective control, we should develop methods that must be strong to compensate for any malfunction that might deviate the system to accomplish its purpose. Therefore, we are considering all possible steering failures, so that we could prevent them from happening. Fault detection and isolation is a system that can be implemented to detect and correct faults. It will work along with the current controller to perform its job. Usually, fault detection is challenging because not only we should know exactly the dynamics of the system, to be able to identify any intruders, but we also should have a clear idea about the faults we are trying to detect. The interface of other subsystems will eventually have an impact on the faulty system if by any chance they are interconnected. Therefore, we also need to know the other surrounded systems.

Fault tolerant control systems-table-01.jpg
Fault tolerant control systems-table-01-2.jpg

P.S: the objective of a fault tolerant control system (FTCS) is to maintain current performances close to the desirable ones and preserve stability conditions in the presence of component and/or instrument faults [1]. That rule should apply to all faulty subsystems. Some of the steering failures are simulated in carsim, such as a flat left tire, a drift by the driver to the right. CarSim predicts the performance of vehicles in response to driver control inputs (steering, throttle, bakes, clutch, and shifting) in a given environment (road geometry, coefficients of friction, wind). By performance, we mean vehicle motions, forces, and moments involved in acceleration, handling, and braking. Just about any test of a vehicle that would be conducted on a test track or road can be simulated [7].

Fig 3: Flat left tire
Fig 4: Vehicle drifts on the right

Most fault tolerant control systems require redundancy in the hardware components to make these systems more reliable. However, the additional parts increase tremendously the overall cost of the system. To develop the analytical redundancy, we first need to understand the steering dynamic [1]. From previous work done through documentation, it was fairly easy to be familiar with the steering dynamics parameters of Cajunbot. A short review of the steering controller of Cajunbot is covered later.

[edit] Faults

Fault is defined as any change in a system that corresponds to abnormal behavior of the system’s regular operation [3]. The abnormal behavior can be generated by component failure such as actuators, sensor or the system itself. Different parts of the system can be corrupted by faults. Hence, faults can be classified according to the location of occurrences and their characteristics. However, in the purpose of the research, we decided to focus on making the steering controller of the cajunbot fault tolerant. Therefore we will not be focus on all the other faults types.

[edit] Actuator Faults

Actuator faults need to be taken of very quickly because actuators are the main components which deliver power into the system and allow the vehicle to move at an acceptable manner. Actuator faults in a vehicle might be a loss of speed, a noisy engine, engine leaks, low pressure tires, improper shifting, and hard steering.

[edit] Sensor Faults

Cajunbot is equipped with lot of sensors, such as radar, environmental sensors, IBEO, Sicks that can fail at any time. In order for the robot to be able to travel a path those sensors need to be operated efficiently and accurately. That means sensors reading should be accurate. Sensor faults can be an incorrect reading, no reading, de-calibration, defection, loose wire, overheating, incorrect inputs and outputs.

[edit] Component Faults

The changes in physical parameters of the systems are related to component faults such as weight, component defined parameters, vehicle dynamics coefficients. They results in a change of the dynamical behavior the controlled system.

[edit] Review of Cajunbot Steering Controller

In order to work with the steering, it is important to understand the current steering controller of the vehicle and all its parameters. The steering of Cajunbot is controlled independently either path planning or speed control. Thus, the goal of the steering controller is to minimize error while following a reasonable path at a reasonable speed. It is assumed that all paths and speeds fall well within range of the vehicles operating abilities. The basic look-ahead method is used to design the steering controller. This method uses the lateral deviation from the path of a virtual sensor point at a specified distance in front of the vehicle as input and seeks to minimize this. The distance from the vehicle’s center of gravity to the virtual point is called the "look-ahead distance." The transfer function from steering angle to lateral acceleration of the virtual sensor point is the following [8]

Fig 5: Lateral Vehicle dynamics

Stability and response are the primary concern to achieve good steering; thus, a lead compensator was implemented. the gain is adjusted until maximum performance is observed. For Cajunbot, the following control law was used[9].

Fault tolerant control systems-formula-02.jpg

The step response of the system is shown below:

Fig 6: steering controller step response

The objective was to develop a steering control system for automatic lane keeping; it is useful to utilize a dynamic model in which the states are in terms of position and orientation error with respect to the road. The dynamics model of the vehicle in terms of error is given in [9]. The state variables are readily available from the combination of GPS and INS system of Cajunbot.

Fig 7: linear vehicle model

[edit] Classification of the basic Fault Tolerant Control Groups

Fault tolerant control systems-fig-08.jpg

Fault tolerant control is divided into two approaches: passive fault tolerant and active fault tolerant [3].

[edit] Passive FTC

A passive fault tolerant control can tolerate faults in a satisfactory manner without any control configuration. The controller for this type of design uses robust control techniques to enable the closed loop system to remain insensitive to faults [3]. In this approach the same controller is used throughout normal case as well as fault cases such that the passive fault tolerant controller is easily implemented [10].Robust control techniques enable the control parameters of the controller to remain fixed, so it tolerates changes of the plant parameters. The control system satisfies its goal under all faulty conditions. Hence, fault tolerance is obtained without changing the controller parameters [5]. Passive FTC deals with a presumed set of system component failures based on the actuator redundancy at the controller design stage. The resulting controller usually has a fixed structure and parameters [1] However, the main disadvantages of a passive FTCS is that as the number of potential failures and the degree of system redundancy increase, controller design could become very ambiguous, and the performance of the resulting controller could become significantly conservative. Moreover, if an unanticipated failure occurs, passive FTC cannot ensure system stability and cannot reach again the nominal performance of the system [1].

Fault tolerant control systems-fig-09.jpg

There are different robust control methods to implement a passive fault tolerant control such as quantitative feedback theory, H∞ optimization method, linear- quadratic- Gaussian method (LQG), μ Synthesis, variable structure control (VSC) [3].

[edit] Active FTC

Research on active fault tolerant control system was mostly motivated by flight control systems for aircrafts. The objective is to incorporate a self repairing procedure to ensure safe landing. Unlike passive FTCS, active FTCS does not rely on fixed controller parameters. Active fault tolerant reacts to the occurrence of faults by applying proper manipulation of redundancies. In many cases, a compromise has to be made to accept a degraded performance in the presence of faults due to limited amount of redundancies. The structure of an active fault tolerant control has been shown in the figure below. An active fault tolerant control is consisted of different subsystem such as controller reconfiguration mechanism, fault detection and diagnosis, and reconfiguration control. All three subsystems must work together within time to accomplish an effective performance of the active FTCS [4].

Fault tolerant control systems-fig-10.jpg

The idea of active fault diagnosis is to disturb the system by auxiliary inputs signals. Faults are detected and isolated based on the signal and signal measurement by using online tests [11]. Fault detection, fault diagnosis/isolation consists of the recognition of which components of system are failed [12]. It is very important to detect any presence of the fault at an early state because appropriate actions can be taken to avoid total failure of the plant [3]. The basic idea of the fault detection/ diagnosis scheme is to compare the expected system behavior against the real observed one [4].The control reconfiguration mechanism will react depending on the response of the Fault detection/isolation module. The goal of this controller is to maintain stability and performance characteristics [13].

A variety of active methods for FTC exist. Due the ability to deal with a larger class of faults, active fault tolerant control methodologies have gained more attention in research than passive ones. The main approaches to implement active fault tolerant are: multiple model control, model predictive control, and pseudo-inverse method, model following approach, Eigen structure assignment, and adaptive control.

[edit] Conclusion and Future work

This report presents the different approaches of fault tolerant control. It also emphasizes on different types of faults that can be present in a given system. Fault tolerant control system is very important for autonomous vehicles as well as human operated vehicles. Modern technological systems rely on redundancy design and sophisticated control system to meet ever- increasing reliability and performance requirements [13]. Future work to be done includes the design of a reliable fault tolerant control system for the steering controller of cajunbot. This design will take advantage of both active and passive fault tolerant control.

[edit] References

  1. 1.0 1.1 1.2 1.3 1.4 1.5 1.6 Francois Bateman, Hassan Noura, Mustapha Ouladsine "Actuators fault diagnosis and tolerant control for an Unmanned Aerial Vehicle" 16th IEEE conference on control applications, Singapore, 1-3 October 2007
  2. 2.0 2.1 2.2 2.3 2.4 Xiang Yu, Xinmin Wang, Kairui Zhao."An autonomous robust fault tolerant" IEEE control system, 2006
  3. 3.0 3.1 3.2 3.3 3.4 3.5 Prasad Pilla , "Thesis presented on fault tolerant control architectures of Aircraft systems" University of Louisiana, Lafayette, Fall 2007
  4. 4.0 4.1 4.2 Jin Jiang "Fault tolerant control system - An introduction overview" ACTA AUTOMATICA SINICA, Vol 31,No 1, January 2005
  5. 5.0 5.1 Blanke, M., M., Kinnaert, J., Lunze, M., Staroswiecki (2006). Diagnosis and Fault Tolerant Control. Springer-Verlag: Berlin, 2nd ed.
  6. IEEE International Symposium on Fault-Tolerant Computing (FTCS) "Fault-tolerant computing"
  7. Mechanical simulation "CARSIM quick start guide"
  8. Arun Lakhotia, Padraic Edgington, Suresh Golconda, Anthony Maida, Pablo Mejia, Gunasekaran Seetharaman "CajunBot-II: An Autonomous Vehicle for the DARPA Urban Challenge" 2007
  9. 9.0 9.1 J. Herpin, A. Fekih, S. Golconda, A. Lakhotia, "Steering Control of the Autonomous Vehicle: CajunBot," AIAA Journal of Aerospace Computing, Information, and Communication (JACIC), vol.4, pp.1134-1142, December 2007.
  10. Afef Fekih, Prasad Pilla, "A Passive Fault Tolerant Control Strategy for the uncertain MIMO Aircraft Model F-18" IEEE 39th Southeastern Symposium on System Theory Mercer University Macon, GA, 31207, March 4-6, 2007
  11. Henrik Neimann "Fault tolerant control based on Active fault diagnosis" 2005 American Control Conference, June8-10, 2005.Porland, OR, USA
  12. P.E Dumont, A. Aitouche, R. Merzouki, M. Bayart, "Fault tolerant control on an electric vehicle"
  13. 13.0 13.1 Dejun Wang, Yuan Chun"Fault tolerant control with actuation reconfiguration" 5th World congress on intelligent control and automation, June 15-19, 2004, Hangzou, P.R china

[edit] See also

Personal tools