Architecture of the VANET Application

With the VANET application, you can test algorithms for the information processing in vehicular networks.

The application implements several components related to the information flow in vehicular networks:

Perceives the environment and implements an error model. These components reside in the folder vehicle at present. See also vanet::ValueTypeIntSensor for an exemplary implementation.
Knowledge base
Organizes and stores various kinds of information. This component resides in the folder knowledge_base. See also vanet::KnowledgeBase.
Information integration
Relates received pieces of information to each other and infers the vehicle's truth or reality of the world. All pieces of information are organised in the knowledge base. The component may include trust and reputation. These components reside in the folder information_integration. See also vanet::EvidenceIntegrator for an exemplary implementation.
Computes trust from own experiences. These components reside in the folder trust. See also vanet::TrustComputer, the interface of these components.
Computes reputation from the recommendations of others. These components are highly related to the trust components. Some related algorithms are sometimes trust and reputation algorithms together. Thus reputation components also reside in the folder trust. See also vanet::ReputationComputer, the interface of these components.
Decides what information is sent to which node in the current situation. This is the right place if you want to implement and attack on the network. These components reside in the folder dissemination. See also vanet::Disseminator, the interface of these components.
The processor controls the interaction between the components and represents the gateway to the simulation core. Currently there is a processor for a vehicle in the folder vehicle: the vanet::VehicleProcessor. A processor for a roadside unit is planed but not implemented yet.

Dividing the information processing algorithms in all these components makes it easy to add and test new algorithms. Finally you can easily compare your algorithms with those of others in the same scenarios.

The shawn::Node and the vanet::VehicleProcessor manage the interaction between these components and connect them to the VANET and the simulation core.

The following section detail the information flow in the VANET application. After that, I give example on how you could implement and test your algorithms with the help of this application.

Information Flow

The VANET application contains algorithms for all parts of the information flow in a vehicular network – for gathering, distributing and organizing information. This section details the interaction of the application's components when processing information. The descriptions and illustrations always take one concrete component (that is, a certain algorithm or model) as an example, although the components can easily be exchanged. This form of description makes it easier to understand the architecture of the application and the interaction between the components.

Gathering Information from the Environment

In a real vehicular network as well as in the simulated one, all information has its origin in the environment, that is, in components that are not part of the vehicle or the vehicular network. Examples might be the driver, a traffic sign, or simply the situational setting of the environment (like ice on the road). A vehicle has input components to receive this information: the sensors.

In Shawn, a reading::Sensor gets information about the environment from a reading::Reading (see session 5 of the tutorial Shawn in 3 days). The reading represents the environment in the simulation. The vanet::ValueTypeIntReading can be filled with environmental events (pieces of information) through a readings file. Vehicles can then talk about these environmental events through the network. So all you have to do, is to create the readings file. This way, you get the information in the network. The Python program (in the sub-folder environment) helps you to spread random events over a SUMO network and save them in a readings file.

The following figure shows the information flow from the readings file into the knowledge base. The task create_environment reads the readings file and saves it in the vanet::ValueTypeIntReading. This must happen before the simulation starts. Then in every simulation round, every vanet::VehicleProcessor polls its vanet::ValueTypeIntSensor once to get the current sensor value. The sensor in turn asks the reading for the current value at the vehicles location. This value is given back to the vehicle, which saves it as an vanet::Observation in the knowledge base.


UML sequence diagram that illustrates the information flow from the readings file through the sensor into the knowledge base.

You can see, the processor controls this process. It triggers the action while the other components only react on these triggers.

Now, some information has found its way in the knowledge base. The next section shows how information from the knowledge base is sent into the vehicular network.

Sending Information in the Vehicular Network

A vehicle can only send what it knows. So sending a message invokes an information flow from the knowledge base into the vehicular network. A vehicle must first gather information before it can set something. Information can be gathered from the environment with the help of sensors (as shown in the previous section) or from the vehicular network. The image below illustrates the information flow.


UML sequence diagram that illustrates the information flow from the knowledge base into the vehicular network.

Again the vanet::VehicleProcessor triggers the process. It invokes a vanet::Disseminator – in the example, the vanet::ObservationDisseminator. A disseminator decides what information to send. Several algorithms have been proposed for this purpose in the literature. For example, you could always send the newest pieces of information, or the most important ones, or a mix of new and old pieces of information. Encapsulating the algorithm in a class makes it possible to exchange it easily per configuration. This way, you can test your vehicular network with different algorithms and even invent your own one.

The disseminator can also model an attack on the network. An attacker would like to send certain messages in the network. A disseminator can control this. Thus if you want to implement a certain attack on the network, you must implement the attack as disseminators in the simulation. For example, the vanet::ModifyingDisseminator implements an attacker that sends good information for a while and switches its behaviour after a configurable simulation round starting its attack. The attack is quite simple then: A certain ratio of messages is manipulated.

A vanet::VehicleProcessor can have several disseminators at once. They are called one after the other during one simulation round. The disseminators can be for different kinds of messages, or for an attack and for usual behaviour.

Gathering Information from the Vehicular Network

Now a message has entered the vehicular network. What happens in nearby vehicles that receive this message? The figure below illustrates that process. It is quite simple: The vanet::VehicleProcessor receives the message from the simulation core and tranforms it in a vanet::Report. A report is simple that part of a message that should be saved in the knowledge base. Then the vehicle processor passes the report to the knowledge base. In addition, it saves that it has met the other node. (This is kept for statistical output.)


UML sequence diagram that illustrates the information flow from the vehicular network into the knowledge base.

So far, this documented introduced the main information flows and the algorithms that process the information. The flow either starts or ends in the knowledge base of the vehicle. This is an important property of the VANET application. You can also see that the processes start at the vanet::VehicleProcessor. This is the gateway to the simulation core. So if you want to gain full control, you simply derive a new processor class from vanet::VehicleProcessor and override the methods as you like.

Integrating Information in the Knowledge Base

This last subsection covers the processing in the knowledge base. It is not another information flow but describes the important algorithm for information integration. That algorithm organises the knowledge base. The figure below shows how it is invoked from the knowledge base.


UML sequence diagram that illustrates the processing in the knowledge base.

The process starts with some incoming information – be it an observation or a report as described in the previous subsections. The knowledge base must now decide whether to keep this new piece of information and where to assign it to. In its internal data structures, the knowledge base relates pieces of information to each other. An observation is associates with a location and a type of information as well as with a time span. In addition, a reported is connected to its sender and its reception time. So an algorithm must decide, how a new piece of information relates to other objects in the knowledge base. The algorithm in the information integrator exactly does this. The algorithm also derives the vehicle's truth or reality from that. This means, for example, if a vehicle receives several message about the same thing just with different observation times, it will assume that the received event was valid over a complete time span, not only at the given points in time. So the vehicle wants to reason about the world with the received observations and reports. This reasoning is done by the information integrator.

Again several algorithms exist in the literature to do this. To make the algorithm exchangeable, it has been encapsulated a the separate class.

How to Extend the VANET Application

Generated on Fri May 24 00:01:56 2013 for Shawn by  doxygen 1.5.6