Medium access methods in communication networks
Ensuring real-time data exchange between NCSs’s components is responsibility of the entire com-munication stack. Communication mediums usually provide limited number of simultaneous medium access for their users. In modern communication networks, medium access constraints are often resolved via various Medium Access Control (MAC) protocols. They define scheduling and collision arbitration strategies in the network. The choice of a medium access strategy is a crucial step because it determines the cost, complexity and data rate in communication system.
MAC protocols can be divided into two groups : Random MAC protocols and sequential MAC protocols. Under a random MAC protocol, every node is allowed to access to medium whenever it has a packet to transmit. If more than one node want to transmit their packets at the same time, an arbitration strategy is used to resolve the packet collision. In the sequential MAC protocols, each user of network accesses to the medium according to a pre-defined sequence.
ALOHA [Abr70] was the first random MAC protocol that was developed by the university of Hawaii for wireless communication. CSMA/CD (Carrier Sense Multiple Access/Collision Detection) which is used in Ethernet and CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) which is used in IEEE802.11 wireless LAN are examples of this group of MAC protocols. These protocols allow real-time medium access arbitrations and they can be used in a network with burst transmission and relativity large data packet. On the other hand, sequential MAC protocols are well-suited for fixed bit-rate and small packets. They provide low latency and bounded delays. TDMA (Time Division Multiplexing Access) is a example of sequential protocols.
In real-time networks, there are several access protocols. Some important protocols are briefly described as follow :
CSMA/CA and CSMA/CD
In the class of CSMA/CA protocols, each message is characterized by a unique priority. Since the shared communication medium can transmit only one message at each instant. Each transmitting node first checks whether the medium is free. If the network is free, then the node can start transmitting. However, it is possible that more than one node detect free communication medium and want to start transmitting at same time. In this situation the transmission of the highest priority message is continued and other messages with lowest priorities are discarded (Figure 5.12). In CSMA/CD access protocol, each node that wishes to transmit a message must initially check whether the medium is free. if the medium is free, the node begins the transmission. It can be possible that two nodes start transmitting simultaneously. In this case, they stop message transmitting and then restart transmitting after a random backoff time.
Token bus was standardized by IEEE standard 802.4. It is the basis of many communication proto-cols such as Profibus, ControlNet [Int99] and MAP [Gro88].
In this protocol, the network nodes construct a virtual ring. The access arbitration is performed by cir-culation of the token between nodes over the virtual ring. A token is passed around the network nodes and only the node possessing the token may transmit. As a consequence, only one node has the token at each moment and it has access to the medium. By this way, collisions are avoided. If a node that takes possession of the token dose not have any information to transmit, it passes the token to its successor. So, each node must know the address of its neighbor in the ring. This protocol guarantees an upper bound on the medium access delay and network expansion can be achieved without a significant drop in performance.
Medium access constraints
When communication medium can only provide limited number of simultaneous medium access channels for its user. As a consequence, only limited number of sensors and/or actuators are allowed to communicate with the controller at each instant k.
In classical control theory a perfect data exchange between system components is assumed. Therefore in NCSs with constraint on communication access, control/FDI design is not only design a classical feedback controller/FDI system. It involves also defining a medium access scheduling strategy. In some works, a Zero order hold (ZOH) at the receiving end (i.e. at the plant’s and controller’s input stages) was assumed [Hri00]. In this case, if an actuator or a sensor fails to access the medium, the most recently updated values stored in ZOH will be used by the plant or controller. An alternative way is ignoring failed sensor or actuator in controller or plant [ZV06]. In this thesis, this alternative way is considered for sensors or actuators which lose their access to the medium.
One of the simplest scheduling strategy is static scheduling. In this scheduling strategy, sensors and ac-tuators access to the medium based on a predefined sequence, termed communication sequence [Bro95]. An LQG design method for NCSs which are subject to medium access constraints was presented in [ZHV05]. The reachability and observability analysis of an NCS with limited communication was stu-died in [ZV06]. In [ICC06] an algorithm for building the command and communication sequence that ensure the reachability and observability of a NCS with medium access constraints was taken into ac-count.
For preserving performance of feedback controller or diagnostic module, sometimes a sensor or actua-tors require immediate attention. So, under static scheduling, NCS may be less robust to unpredictable disturbances. In addition, by using a static scheduling, a global timer is needed to synchronize all sensors and actuators. As a solution to these problems, dynamic medium access scheduling was proposed. By using this scheduling, access to medium is determined on-line based on control/FDI needs. The works of [WY01, WYB02] studied a dynamic scheduling. They proposed the MEF-TOD (Maximum Error First, Try once Discard) policy to stabilize a general NCS.
Quantization and feedback control in NCS
Quantization has been discussed in the context of digital control and signal processing for over thirty years. An analog signal must be quantized before transmitting via a digital communication medium. A quantizer acts as a functional that maps a real-valued function into a piecewise constant function taking on a finite set of values. Normally, quantizers round off the signals uniformly with a constant step. It can help reduce the size of packets and data rate in communication medium. It is possible to increase sensitivity of quantizer as the system state approaches to zero [BL00]. It showed that this policy of quantization can stabilize a linear system by a linear time-invariant feedback controller if the system is stabilized by the feedback controller without quantization. Normally, by increasing control signals’s update frequency, disturbance rejection abilities will be impro-ved. Whereas, increasing their quantization precision improves the steady state performance. But in case of limited bandwidth, increasing quantization precision necessities the reduction of update frequency and vice versa. [BGc10] proposed a strategy to choose online the update rate and the quantization precision level.
Some detailed discussions about quantization problem in NCS can be found in [SRAB04, LB04, Sal05, BTX03].
Regardless of the network architecture/medium/protocol the outcome of a NCS is affected to other factors. For details, the interested reader is referred, for instance, to the textbook [IF02] and papers [LW08, RMY09].
Model-based fault detection
Fault free condition in a controlled plant can not be guaranteed and faults may occur in the system anytime. For instance, in a process industry valves and hoses can leak, bearings can jam, sensors can be in error and so on. They may cause the degradation in the performance of the plant which affect the operation cost and quality of final product. For example [ISSC09] showed that industries lose billions of dollars every year due to plant faults. Fault detection and Isolation(FDI) is the primary stage of fault-tolerant control systems. The interested reader is referred to [Ise05, CP99, Din08, Ise11] for some basic terms used in FDI literatures such as faults, failure, disturbances, uncertainties,… . Different approaches for fault detection have been developed. They can be classified into the following categories [Din08] :
– Hardware redundancy based FD.
– Plausibility test.
– Signal-based FD.
– Model-based FD.
The model-based approach to fault diagnosis in industrial processes has been receiving remarkable at-tention since 70s. Model-based fault diagnosis can be defined as detection, isolation and characterization of faults in components of a system from the comparison of the system’s available measurements, with a priori information represented by the system mathematical model [CP99].
The basic idea of model based FD system for a process is illustrated by Figure 2.10. The mathema-tical model of the process is running simultaneously with physical process. It is driven by process’s inputs and it estimates process’s outputs (without considering disturbances). In fault free condition, the co-called residual signal which is the difference between the estimated outputs and measured outputs, should be zero. If there is a fault in the process, the residual will be deviated from zero. The procedure of creating the estimated outputs and building the residual for extracting fault symptoms from the system is called residual generation. The process which is used to generate residuals is called residual generator. The second step in the procedure of fault detection is Decision Making. In this step, a decision rule is applied to determine if any fault is occurred. A decision rule may consist of a simple threshold test on the instantaneous value of residual. This threshold may be a fixed or a time-varying value.
Table of contents :
1.1 Why Networked control systems
1.2 Safe-Necs(Safe-Networked Control Systems) Project
1.3 Objectives and motivations
1.4 Thesis outline and contributions
2 Networked Control Systems
2.1 A brief history of NCSs
2.2 NCSs Categories
2.3 Medium access methods in communication networks
2.3.1 CSMA/CA and CSMA/CD
2.4 Fundamental issues in NCSs
2.4.1 Network-induced Delay
2.4.2 Packet dropout
2.4.3 Medium access constraints
2.4.4 Quantization and feedback control in NCS
2.5 Model-based fault detection
2.5.1 Fault detection of Networked control systems
126.96.36.199 Fault detection in NCSs with network delays
188.8.131.52 Fault detection of NCSs with packet losses
3 FDI with limited communication : Fixed scheduling case
3.1 Communication limitation modeling
3.2 Observability and Reachability of extended plant
3.3 Fault detection
3.3.1 Observer-based approach
3.3.2 FD subject to access constraints and Packet dropout
3.4 FD in distributed system with communication constraints
3.5 Design of structured residual sets in NCSs subject to random packet dropouts
3.5.1 Fault isolation
3.5.2 Illustrative example
3.6 FDI of NCSs subject to random packet dropout and medium access constraints
3.6.1 Design of robust residual generators
3.6.2 Illustrative example
4 Communication sequence design in NCS with communication constraints : a graphic approach
4.1 Structured system
4.2 Graphic representation of linear structured systems
4.2.1 Directed graph
184.108.40.206 Generic controllability and generic observability
4.2.2 Dynamic bipartite graph
4.2.3 Generic reachability
4.3 Communication sequence design, observability
4.4 Communication sequence design, reachability
5 Fault detection in limited communication with dynamic scheduling
5.1 Problem formulation
5.2 Parity space-based residual generation and semi-online scheduling
5.2.1 Offline communication sequence design
5.2.2 Sequence selection
220.127.116.11 Multi Sequential Probability Ratio Test(MSPRT)
5.2.3 Fault detection
5.3 Observer-based residual generation and semi-online scheduling
5.4 Online scheduling in CAN network by means of hybrid priority
5.4.1 CAN network
5.4.2 Hybrid priority
5.4.3 Hybrid priority with diagnostic objective
6 Drone Application
6.1 Drone presentation
6.1.1 Architecture of prototype of Safe-NECS project
6.1.2 Drone movements
6.1.3 Attitude representation
18.104.22.168 Euler angles
6.1.4 Drone sensors
6.1.5 Mechanical Model and attitude estimation
6.2 Fault tolerant control
6.2.1 Reconfiguration in case of critical failures
6.2.2 Actuator and network faults and its effects
6.2.3 Fault tolerance control module
22.214.171.124 First strategy
126.96.36.199 Second strategy
188.8.131.52 Third strategy
184.108.40.206 Fourth strategy
6.3 Semi-online scheduling and fault detection
6.3.1 Residual generation
6.3.2 residual classification
6.3.3 Sensor access scheduling
6.3.4 Simulation results
7 Conclusion and future works