WO2008095518A1

WO2008095518A1 - Use of a distributed diagnostic architecture in autosar

Info

Publication number: WO2008095518A1
Application number: PCT/EP2007/009922
Authority: WO
Inventors: Michael KÖHLER; Tim Schluesener; Thomas Schreiber; Martin Voigt; Stephen Wilde
Original assignee: Daimler Ag; Fischer, Andre
Priority date: 2007-02-06
Filing date: 2007-11-16
Publication date: 2008-08-14
Also published as: DE102007006614A1

Abstract

The invention relates to an E/E architecture, the control devices and applications of which have implemented the AUTOSAR standard, and to a multi-agent system which is developed by means of an object-oriented program language and the functional version of which can be integrated into the RTE of AUTOSAR. For this purpose, the components involved in the E/E architecture each have an AUTOSAR interface onto which interface agents are placed for each involved control device of the multi-agent system. The multi-agent system can then use its agents to access, via said interface agents, the control devices and to use the communication connections thereof. The multi-agent system is used for the distributed diagnostics of the E/E architecture and relates to the diagnostics of functions and hardware components. The used diagnostic agents of the multi-agent system permit an error in the system to be located down to the smallest exchangeable unit. The distributed diagnostic system comprises both onboard agents and also offboard agents for the diagnostics.

Description

Application of a Distributed Diagnostic Architecture in AUTOSAR

The invention relates to an electrical / electronic architecture (E / E architecture), a distributed diagnostic system and a multi-vehicle agent system for motor vehicles. With the technique of multimedia agents, partial diagnoses of single agents performing the distributed diagnostic tasks in the scope of the system to be diagnosed are combined to form a diagnostic result. A new E / E architecture and software architecture enables the implementation of the multi-agent system-based automotive diagnostic system.

The background for the use of the new technology are the increasing complexity of distributed systems in motor vehicles in recent years. The degree of crosslinking of sensors, actuators, control units and software is constantly increasing. This enables completely new functions in the vehicle, often interacting with several control units, sensors, actuators and programs in their creation. The diagnostic quality of diagnostic systems must keep up with the increasingly distributed character of the E / E architecture. In addition, there is the necessity that the diagnostic system must be flexible and at the same time robust against an increasing number of equipment variances of the vehicle electrical system. With centralized diagnostic systems, these tasks can only be mastered with great effort. The increasingly distributed E / E architecture in motor vehicles also leads to increased integration effort in the cooperation between automotive suppliers and automobile manufacturers in order to be able to provide the compatibility of the individual systems necessary for integration. In this context, automobile manufacturers and automotive suppliers have joined forces to form a consortium to create common interfaces for the software architectures. This cooperation has created a common interface standard for ECU software. This interface standard is called AUTOSAR. An overview of the most important features and possibilities of AUTOSAR is described in the article by Matthias Wernicke and Jochen Rein: "Development of ECU software according to AUTOSAR", Elektronik Informations 11, 2006, Page 78-80 The heart of the AUTOSAR standard is a Run Time Environment (RTE), which organizes the communication exchange and data exchange and forms the basis for a uniform application interface for individual application software C ++ applications, more program code support is expected to follow.

With the help of multi-agent systems, various tasks in distributed technical systems can be solved. Multiagent systems have been described in numerous publications. Michael Wooldridge gives a theoretical overview with "An Introduction to MultiAgentSystems. "John Wiley & Sons Ltd., 2002. ISBN 047149691X An overview of applications in which agent systems could have a major impact is given by Birgit Burmeister, Asaneh Haddadi, and Guido Matylis in Applications of Multi Agent Systems in Traffic and Transportation. "In IEEE Proceedings of Software Engineering 144 (1997) 1, pp. 51-60, this study cites, in particular, applications in the field of traffic control systems and fleet management of vehicle fleets.

In the field of diagnostic systems for motor vehicles is in Haus der Technik textbook: "Modern electronics in motor vehicles", Bernard Baker (ed.) And 62 co-authors, expert Verlag 2006, ISBN 10: 3-8169-2575-8, p 126-141 by Michael Fischer, Hans Christian Reuss, Stephen Wilde, Michael Köhler a contribution "multi-agent based architecture for the distributed diagnosis in the motor vehicle" revealed. It is reported about a multi-agent diagnosis for a rear lid remote control of a motor vehicle. The control units and actuators for the rear lid remote control and the associated communication system are modeled and simulated using a Matlab Simulink model. The Java based multi-agent system is tested with the simulation model. The invention disclosed herein is based on this prior art.

For the realization of a multi-agent system in the diagnosis of distributed architectures in motor vehicles, the following obstacles arise with the known prior art.

Development tools that can be used to create multiagent systems are usually based on the application of object-oriented programming languages such as C ++ or Java. A well-known agent platform is eg the Java based platform JADE - Java Agent DEvelopment Framework. Jade is available as a freeware from Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA. Another toolkit for creating agents is provided by Acronymics, Inc., 1301 West 8th St., # 28, Mesa, Arizona 85201-3841, sold under the name AGENTBUILDER - "An Integrated Toolkit for Constructing Intelligent Software Agents."AGENTBUILDER's Runtime Environment supports Java, C, and C ++.

Distributed E / E architectures in motor vehicles use machine-programmed and object-oriented software only due to the very limited resources available to the microcontrollers of the participating ECUs. Control unit computers in motor vehicles are usually programmed with the current development tools in C and therefore contain compiled C code. Object-oriented programming languages such as C ++ or Java can not be used in previously known E / E architectures for motor vehicles. An implementation of the simulated multi-agent system according to Fischer, Wilde et al. in a prior art E / E architecture of a motor vehicle is therefore not possible or only with great effort.

It has therefore been an object of this invention to find a way in which multi-agent systems can be implemented in E / E motor vehicle architectures.

The solution succeeds with an E / E architecture and with a multi-agent system for distributed diagnostics in motor vehicles with the features of the independent claims. Further possible embodiments of the invention are disclosed in the subclaims, the graphs and the following description.

For this purpose it is envisaged that an E / E architecture whose ECUs and applications have implemented the AUTOSAR standard, and a multi-agent system developed by means of an object-oriented programming language whose executable version can be integrated into the RTE of AUTOSAR, be applied. The multi-agent system, which is implemented in the form of an agent platform and various agents, diagnoses the hardware components as well as the functions within the E / E architecture in motor vehicles.

AUTOSAR, whose standard allows the use of object-oriented programming languages, allows multi-agent systems to be developed and integrated into the distributed E / E architecture of a motor vehicle. The multi-agent system is preferably programmed in Java and is used e.g. Using the java2C ++ tool from www.programics.com, converted to C ++ code supported by AUTOSAR's RTE. Alternatively, the multi-agent system may also be used with a C ++ oriented development platform, such as a. AGENTBUILDER be built. A conversion from Java code to C ++ code can then be omitted. Should the AUTOSAR standard also support Java in the future, multi-agent systems programmed in Java can be directly implemented without conversion to C ++.

In order to implement the multi-agent system in the E / E architecture, each of the components involved in the E / E architecture has implemented a realization of the agent platform on which the agents of the multi-agent system are set up. The agent platform enables the use of agents within the multi-agent system. The components also each have an AUTOSAR interface to which a hardware agent is set up by the multi-agent system for each control unit involved. The multi-agent system, with its agents, can access the ECUs and their input and output signals, such as controls or sensor data, internal states, or controller fault memory via these hardware agents, thereby enabling diagnostics in the multi-agent system. With the aid of the multi-agent system, a distributed diagnosis is implemented in the aforementioned E / E architecture. The distributed diagnostic system distinguishes between diagnostics in motor vehicles between functions, subfunctions and hardware components such as control units, actuators or sensors. The main advantages that can be achieved with this are the improved mapping of vehicle functions to the E / E architecture. Agents are ideal for encapsulating and summarizing functions or sub-functions that run across multiple ECUs. In other words, distributed (sub) functions can be mapped and summarized very well with an agent. The reason lies in the abstraction of the hardware-related E / E architecture in the aforementioned functions, sub-functions and components. This has particular advantages for the diagnosis, because for every perceptible vehicle malfunction or for every perceptible fault pattern, it is possible to use special diagnostic agents for the diagnosis of precisely this observed behavior. If the corresponding behavior occurs, the respective diagnostic agent becomes active and creates a diagnosis with respect to the error, either alone or in conjunction with other agents.

Based on the above, in the multi-agent system for distributed diagnosis, agents for function-oriented diagnosis and for component / component-oriented diagnosis are used and set up on the agent platform. Each diagnostic agent is assigned a technical system scope to be observed by it according to this classification. The diagnostics agent creates a diagnosis for this subsystem. The diagnostic agent uses programmed knowledge about the fault setting conditions in the subsystem he is observing, but can also make use of existing self-diagnostic routines of the control units, actuators or sensors. In addition, diagnostic agents use specially for the diagnosis of functions knowledge about dependencies of the distributed functions to neighboring subsystems.

An important basic principle within the multi-agent system is cooperation. The separation into task areas of the agents leads to the fact that each agent can make at first only analyzes and statements to the system extent observed by it. For this he uses the methods of state recognition and self-diagnosis implemented in this system scope. If an error is found during this process, which was actually caused by another subsystem, the system behavior of the influencing subsystems must also be investigated in order to avoid possible misstatements due to the distribution. Since their behavior is in turn monitored by other diagnostic agents, further partial diagnoses are available for these subsystems. In order to arrive at an overall diagnosis of the system, cooperation among the agents of a multi-agent system is required. In order to be able to create this overall diagnosis, the participating agents communicate with each other and try to resolve existing inconsistencies from the agent-specific partial considerations, including the global knowledge such as error dependencies. The communication process is continued in this case with the involvement of further subsystems and other agents until a consistent and complete explanation for the diagnosis causing misconduct has been found.

In order to enable the communication of the specific agents with each other and in particular beyond the limits of the control units, existing physical communication links between the subsystems are used. For this purpose, in each participating control unit a

Be implemented communication interface. It is implemented by a platform agent who is integrated into the multi-site system, which connects the agent platforms of the individual ECUs and is based on the runtime environment of the AUTOSAR standard. The platform agent can be used, for example, to exchange messages between the agent platforms of the subsystems for agent diagnostics.

An embodiment of the invention will be explained in more detail below with reference to graphical representations.

Showing:

1 is a schematic E / E architecture for a motor vehicle with diagnostic interface,

2 shows an E / E architecture according to the invention, FIG. 3 shows a schematic representation for the implementation of a multi-agent system in an AUTOSAR environment,

4 shows a schematic structure of a function agent, FIG. 5 shows a possible hierarchical structure for a multiagent system.

FIG. 1 shows by way of example an on-board network structure or in other words an E / E architecture for a motor vehicle. Such on-board network structures are currently used in motor vehicles. The control units installed in the motor vehicle, whose number in modern vehicles easily reaches the order of 60-80 pieces, are in communication with each other via data buses. A common data bus in motor vehicles here is the so-called CAN bus (for Control Area Network). Each of the installed ECUs ECU1, ECU2, ECU3, etc. in the motor vehicle has the ability to self-diagnose in addition to the communication interfaces. As part of the self-diagnosis of the control units, errors detected in the control units with the aid of the diagnostic routine are codifi- zierter form as so-called error codes from the ECU software in specially reserved memory areas, so-called error memory, written. In the schematic representation of FIG. 1, these reserved, non-volatile memory areas are designated FS (for error memory). In this case, an onboard diagnostic program can be implemented in one or more of the on-board network control devices. The E / E architecture provides a diagnostic interface 2 for communication and for data exchange between a diagnostic tester in the workshop and the control units installed in the motor vehicle. This interface makes it possible to connect an external diagnostic tester 1. A gateway converts the communication protocol of the diagnostic tester to the bus protocol of the E / E architecture. This situation is standard today in motor vehicles and workshops.

Starting from the aforementioned E / E architecture, FIG. 2 shows a schematic representation of the inventive E / E architecture for a multi-agent system (MAS). To E / E architectures as shown schematically in Fig. 1, there is with AUTOSAR a new interface standard, which provides a uniform interface to the base software of the control unit for the application developments APPl, APP2. Among other things, data such as the contents of the fault memory FS, various services of the control devices, eg for memory access or for diagnosis, can be used via this interface, but also the communication interfaces of the respective bus system can be accessed. AUTOSAR hereby assumes the classic role of middleware. This means that different applications can use an interface that is uniform in the direction of the application, regardless of the hardware of the E / E architecture. The example of FIG. 2 is based on a CAN bus for the communication. In order to be able to use the communication via CAN bus, the AUTOSAR middleware is based on a CAN transport layer consisting of the CAN transport protocol, the CAN controller and the CAN transceiver. For a Flexray, Most, Bytefly or LIN bus, corresponding bus-specific transport layers are used. However, the application layer will always be based on the middleware AUTOSAR, which always offers the same standard for the application layer.

According to the invention, a multi-agent system is implemented in the application layer. For this purpose, as shown in FIG. 3, in each control unit ECU1, ECU2 involved, an agent platform 31 is set up and implemented on the AUTOSAR RTE. Together with the diagnosis agents 32 and the interface agents 33 and 34, they form the multi-agent system. The diagnosis agents 32 are responsible for the distributed diagnosis in the motor vehicle. Each diagnostic agent represents a subsystem in the E / E architecture, ie a (sub-) function or a component, in the sense of distributed diagnostics. In order to be able to create a diagnosis for the respective subsystem, the diagnostic agents need information (current signal values, internal states, possibly self-diagnoses, etc.) about the state of the subsystem to be diagnosed. This information is provided by the hardware agent 33. In addition, it allows the storage of the diagnostic result in the fault memory of the control unit. The platform agent 34 establishes an interface to the transport layer of the control unit via the AUTOSAR RTE. This allows the agents of the multi-agent system to communicate and cooperate across the physical device boundaries and independently of the underlying physical bus. With this invention, a multi-agent system for the distributed diagnosis of motor vehicles is implemented. Single agents perform distributed diagnostic tasks in the overall system, which contribute to solving the overall diagnosis by cooperating with the other agents. The modularity of the diagnostic system saves costs through scalability and reusability of the agents. New wiring systems variants or E / E architecture additions can be made diagnosable at any time by adding new agents to the diagnostic system.

A schematic, typical structure of an agent for diagnosing a function is shown in FIG. Still below, component-oriented agents are discussed. Such a function agent is particularly suitable as a diagnostic agent for a distributed diagnostic system. The function agent intelligently combines existing distributed functions in control units according to their task and the diagnostic competence derived therefrom. In order to be able to create a diagnosis for the function, the function agent uses methods for state detection and condition monitoring, shown in FIG. 4 with the function block 41. If an unauthorized deviation is detected with this state machine, a local diagnosis creation is started and a corresponding error entry is made via an interface 44 in a fault memory. Subsequently, error dependencies stored in the knowledge of the function agent are identified to other technical subsystems and resolved in cooperation with other agents. FIG. 4 shows this behavior with the function block 42. For cooperation with other agents, each diagnostic agent involved in the multi-agent system offers a corresponding interface 43 to the agent platform. This interface can be used to receive messages from the diagnostics agent via the agent platform and the connected platform agents to agent platforms on other control units are exchanged. The results of the cooperation are also entered in the error memory. The diagnostic result present in the error memory can then be output. If a diagnostic tester is connected, we recommend outputting the diagnostic result on the display of the tester.

As a rule, a plurality of dependent components are controlled by a control device, which communicate with the control device in a hierarchical subordination. For an agent system and especially for a diagnostic system based on a multi-agent system, it is expedient to use hierarchical communication structures when setting up the agent system. It has already been mentioned that a function agent should be based as far as possible on a function. Since functions often interact equally with one another in a motor vehicle, function agents should, if possible, be able to interact with one another as equally as possible. In accordance with the communicative dependencies on the functions, component agents should, as far as possible, only interact with the function agents assigned to them and have no possibility of interaction with one another. The corresponding multi-agent system, which consists of function agents and component agents, should have a hierarchical structure, wherein the component agents are hierarchically assigned to the higher-level function agents. This situation is shown in Fig. 5.

With an E / E architecture according to the invention with AUTOSAR standard and agent platform on the control units as disclosed here and starting from a multi-agent system such as, for example, simulation in the initially named publication by Fischer, Reuss, Wilde, Köhler: "Multi Agent Said architecture for distributed diagnostics in motor vehicles "has been described, can be implemented in the further described novel and inventive diagnostic system for motor vehicles and diagnostic tester.

The diagnostic system is based on a multi-agent system. Agents are used for onboard diagnostics in the automotive E / E architecture, and agents for offboard diagnostics are deployed with a diagnostic tester. Furthermore, in the totality of the multi-agent system, the diagnostic agents are realized both as function agents and as component agents. Thus, the two perspectives on the technical system, ie components as well as functions, can be used for the overall diagnosis. The component agents represent the hardware-oriented view of the diagnostic problem and can detect and isolate faults in one's own component. In contrast, function agents are used to detect and classify errors in the functional sequence. They embody the function orientation.

Each of these agents performs a specific task in the diagnostic system. For this purpose, each agent has a specific knowledge about its component or its function and can create a diagnosis specifically for this component or function.

A diagnostic procedure is then designed, for example, as follows.

In order to diagnose errors by the diagnosis multi-agent system, the diagnosis is carried out in three steps. In a first step, first onboard local diagnoses are created, ie each diagnostic agent in the vehicle creates a diagnosis for his or her task only on the basis of their own knowledge. and the associated technical subsystem.

These local diagnoses are aggregated in a second step into a global diagnosis statement by evaluating and calculating the partial diagnoses using the function agents. The basis for this is cooperation and communication between the agents. As part of this cooperation, the multi-agent system resolves error dependencies.

In an optional third step, the global diagnostic results can be refined and supplemented by offboard agents and their knowledge. This is the case when the onboard agents can not make a clear diagnostic statement because e.g. Insufficient information can not be used to calculate a clear error image. With the help of test steps, additional information is obtained by the customer or the workshop personnel, who are used for the evaluation of the already determined diagnostic results and thus narrow down and localize the source of the error.

The creation of local onboard diagnostics is done in two steps. First, an error detection is performed and then made an error limitation.

The error detection is based on the monitoring of the normal behavior of the subsystem to be diagnosed. The respective diagnostic agent monitors corresponding signals, internal variables, limit values and time constants as well as timeouts in the subsystem. The agent-specific knowledge of the subsystem observed by the agent is programmed in the form of diagnostic knowledge by the diagnosis author or diagnosis expert. After a misconduct has been detected, the error is further limited and classified. The diagnostic multi-agent system presented here uses error trees that were also programmed by the diagnostics expert. In principle, however, other classification methods can also be used, such as signal models, Bayesian networks or case-based closure. As a result of the error classification, a defined error object is generated which describes the error and is stored in the error memory. In addition to the error, the error object also contains the error source and a timestamp of when the error occurred.

The procedure described above for creating a local diagnosis is carried out in all diagnostic agents, so that at the end of the first step each agent provides a diagnosis of its observed subsystem. This is the prerequisite for the subsequent step of creating the global diagnosis.

To be able to diagnose the entire system, the agents must cooperate with each other. The cooperation allows to resolve the error dependencies as well as existing logical conflicts caused by the limited views of the local diagnoses. The resolution of the error dependencies is based on the exchange of knowledge between the agents on the presence of diagnosed local errors and their dependence on each other. The prerequisite for this is that corresponding dependency rules between the errors of the subsystem had previously been defined and these error dependencies were modeled as global knowledge. The function agents then determine additional diagnostic agents along the error dependencies that have potential causes for their own functional errors can. The error dependencies are for this purpose in the form Error causes imply consequential errors in the global knowledge of the respective diagnostic agent. Only the dependencies in agent knowledge that are relevant for the respective subsystem are modeled.

After the agents have been identified for further communication, requests are made to them. A reply to a request by the communication partner is generated on the basis of the error entries in the respective error memories. If the requested error exists in the error memory of the subsystem, this is confirmed in the response, while unrecognized but requested errors are declared as not present. Each error object has a special error status attribute for this purpose, which expresses the status of the error. Due to the error status attribute, the response is evaluated in the requesting function agent.

The evaluation of the received responses is carried out together with the evaluation of the error-dependency rules by the function agents. Since the dependency rules represent cause-and-effect relationships, they form the basis for finding causes in the answers. The evaluation is carried out by means of two functions. A confirmation function is used to determine if the malfunction is explained by the reason in the answer. A satisfiability function is used to check whether the functional error can no longer be confirmed by any cause. If the malfunction can not be explained by any onboard available causes, it may be necessary to continue with an external diagnostic tester. If a malfunction is confirmed with a cause, the cause is stored in the fault memory. In the error memory after the global diagnosis thus in addition to the detected malfunction always its diagnosed causes are stored.

As already described above, a multi- agent diagnostic system opens up further possibilities in the off- board area as well. With special agents for this offboard area of the diagnostic system, further checks can be carried out that are not possible with the onboard diagnostic system. In the workshop, the troubleshooting can be continued with special tests and with additional functional sequences. The employee in the workshop carries out a predetermined functional test, which is tuned to the observed customer's symptom, monitors its passage and enters the results of his observation into the diagnostic system. These are forwarded by an offboard-side agent to the other on-board and off-board agents of the multi-agent system and then evaluated. In this way, the multi-agent system can further restrict the diagnostic result by including further information from the workshop tests.

Advantageously, the offboard agents are designed as interactive communication agents who ask questions in the form of dialogues to the user of the diagnostic system and suggest tests that the user answers after appropriate verification and enters the information thus determined using the interactive communication agent in the diagnostic system. The information thus obtained is evaluated using diagnostic rules programmed in the off-board area and corresponding diagnostic results are generated. Existing onboard diagnostic results are displayed in this involved in the evaluation process so that the offboard diagnostics extend the already existing onboard results and thereby achieve a higher diagnostic depth.

In conjunction with the offboard diagnostics, remedial measures are also defined, which should be carried out by the workshop personnel to eliminate the error. These corrective measures are output to the workshop personnel together with the diagnostics from the off-board area. If no unambiguous diagnosis is possible using the offboard agents, further measures must be taken in the off-board area, but these are no longer part of the multi-agent system. These include, above all, the carrying out of test drives and the use of additional information systems with further expert knowledge of experts, e.g. from design databases or from long-term studies on failure probabilities and wear.

The use of a distributed architecture for the diagnosis of a distributed system such as the motor vehicle brings a number of advantages. Thus, by using an agent system, a high variability of the diagnostic system is achieved. The fact that each agent is understood as a single element in the diagnostic system, a major contribution to the extensibility and reusability of the diagnostic system in various series and variants of motor vehicles is achieved, which in turn can lead to cost savings.

The diagnosis of new equipment variants can be made very easy by adding new diagnostic agents. The internal structure of the agents also allows the extension of existing agents with additional diagnostic knowledge in the form of new diagnostic rules or diagnostic facts. In addition, the multi-agent system is designed so robust that individual agents can be released from the network without endangering the diagnostic statements of other agents. Even under these conditions, partial diagnoses are created by the remaining agents and a knowledge exchange in the direction of overall diagnosis is carried out. The confinement of the diagnosis to a replaceable unit may be hampered if the cause of the error lies in the subsystem which the removed or missing agent should observe.

The organization and the processes for creating the diagnostic system can also be supported by the agent solution, since agents can be provided by different diagnosticians or by different suppliers and these agents are able to work together through common interfaces. This allows the knowledge from different sources - developers, suppliers, diagnostic auto- to combine with each other, without loss of communication occur.

Due to the component-oriented agents, the previous component-oriented diagnosis can be completely simulated via error codes. In addition, a higher level of diagnostic depth can be achieved by not only taking component diagnoses but also function-oriented diagnostic statements from function agents. The use of function agents in diagnostics enables the customer's view to be included in the diagnosis. Customer-perceived errors are usually function-oriented. A diagnostic system, which is constructed as a multi-agent system, in which some of the agents are functional agents and part of the agents are component agents, opens up the possibility, based on the cooperation between the agents, of making an error based on the customer complaint Finally, to substantiate the interactions of the functional agents and component agents except for a replaceable unit. Thus, the target of an integrated diagnosis of the customer complaint to establish the most direct connection to a smallest interchangeable unit by the diagnosis system according to the invention by means of multi-agent system achieved elegantly and almost effortlessly. The fineness of the component agents should be adapted only to the fineness of the smallest exchangeable components.

Claims

claims

1. E / E architecture, especially for motor vehicles, in which several control devices and several electrical or electronic components with each other in

Communication connection are, characterized in that at least the control units have a uniform application interface, in particular via an application interface according to the AUTOSAR standard, and that an agent platform is applied to the application interface of each control device.

2. E / E architecture according to claim 1, characterized in that the architecture has an onboard portion in the vehicle and an offboard portion outside the vehicle, which can be connected to one another via an interface.

3. E / E architecture according to claim 1 or 2, characterized in that the communication link as a physical bus using a CAN bus, a Byteflight-, a MOST- a LIN bus, a Flexray BUS or a combination of these bus systems.

4. Multi-agent system with functional agents and component agents for the diagnosis of distributed E / E architectures, in particular for E / E architectures according to Claim 1, wherein the agents communicate with each other via an agent platform, characterized in that the causal error is located in a replaceable unit of the E / E architecture via the function agents and the component agents.

5. Multi-agent system according to claim 4, characterized in that it consists of onboard-side agents and offboard-side agents.

A multi-agent system according to claim 4 or 5, characterized in that individual agents interface with E / E architectures of claim 1 to exchange information between the multi-agent system and the E / E architectures.

7. Multiagent system according to one of claims 4 to 6, characterized in that it is constructed with an object-oriented programming language.

8. Multi-agent system according to claim 7, that the programming language is Java or C ++.

9. multi-agent system according to one of claims 4 to 8, characterized in that each agent has both functional or component-specific knowledge as well as global knowledge of the functional and component dependencies of the E / E architecture has.

10. Diagnostic system for an E / E architecture of a motor vehicle, in which a diagnosis of the E / E architecture can be carried out with a multi-agent system of function agents and component agents, characterized in that at least a part of the agents of the Multi-agent system are implemented onboard in the E / E architecture of the motor vehicle.

11. Diagnostic system according to claim 10, characterized in that the implementation in the control unit of the E / E architecture via a uniform application interface, in particular via an application interface according to the AUTOSAR standard.

12. Diagnostic system according to claim 10 or 11, characterized in that the individual agents communicate with each other via an agent platform.

13. Diagnostic system according to one of claims 10 to 12, characterized in that a part of the offboard agent is implemented outside the E / E architecture of the motor vehicle, in particular in a diagnostic tester.

14. Diagnostic system according to claim 13, characterized in that at least a part of the offboard implemented agents is designed as an interactive agent.

15. Diagnostic system according to one of claims 10 to 14, characterized in that it is expandable with additional agents.

16. Diagnostic system according to one of claims 10 to 14, characterized in that agents can be removed without the running of the system is impaired.

17. Diagnostic system according to one of claims 10 to 16, characterized in that the structure of the multi-agent system images the component agents on the exchangeable units of the E / E architecture.

18. Diagnostic system according to one of claims 10 to 16, characterized in that the structure of the multi-agent system maps the function agents on the functions, sub-functions or distributed functions of the E / E architecture.