Today s automobiles consist of an increasing number of interconnected electronic devices so-called e

Advances in Software Engineering Volume 2012 (2012), 4 star hotel barcelona Article ID 971430, 15 pages doi:10.1155/2012/971430 Research Article A Multi-Layered 4 star hotel barcelona Control Approach for Self-Adaptation in Automotive Embedded Systems Marc Zeller  and Christian Prehofer
Copyright © 2012 Marc Zeller and Christian Prehofer. This is an open access article distributed under the Creative Commons Attribution License , which permits unrestricted 4 star hotel barcelona use, distribution, and reproduction 4 star hotel barcelona in any medium, provided the original work is properly cited. Abstract
There has been considerable work on self-adaptive systems which can reconfigure their software 4 star hotel barcelona configuration 4 star hotel barcelona at runtime [ 1 3 ]. However, applying these techniques to networked, embedded systems poses several new problems due to limitations and reliability requirements of embedded systems [ 4 ]. In particular, we focus on automotive embedded systems, where the main constraints are (i) limited memory resources, (ii) heterogeneous hardware platforms, (iii) different subnetworks connected 4 star hotel barcelona by a gateway, (iv) various requirements of different functionalities, 4 star hotel barcelona and (v) high demand on safety and reliability.
Today s automobiles consist of an increasing number of interconnected electronic devices so-called electronic control units (ECUs) which realize most functionalities of the car by software. This networked embedded system 4 star hotel barcelona keeps the vehicle running by controlling the engine and the breaks, provides active safety features (e.g., antilock breaking system), makes driving more convenient, and entertains the passengers with a large number of information and comfort services (e.g., air conditioning, audio player). Especially modern driver assistance systems, which distribute their functionality over several components, increase the complexity 4 star hotel barcelona of today s vehicular embedded systems enormously. Managing nowadays vehicle 4 star hotel barcelona software systems means managing over 2,000 software components, running on up to 100 ECUs [ 5 ].
Enhancing automotive embedded 4 star hotel barcelona systems with self-adaptation provides a promising solution for the current challenges in automotive embedded systems [ 6 ]. So-called self- properties, like self-configuration, self-healing, self-optimization or self-protection [ 7 ] improve the scalability, robustness, and flexibility of the system [ 8 ]. With the size of automotive systems, it becomes difficult to calculate all configurations and all failure cases in advance. Hence, the adaptation of the system may have to be calculated during runtime. In this work, we focus on the adaptation control of the system after the breakdown of a hardware platform. The actual reconfiguration of the system itself is not the focus here and can, for instance, be performed during a (partial) restart of the system.
For realizing systems with these self-managing capabilities, a control component is needed [ 9 ]. This external component supervises the system and initiates the adaptation during runtime using closed feedback-loops (so-called control loops) [ 10 ]. In the autonomic computing (AC) paradigm, the elements of the system are managed by control loops based on the so-called MAPE-K cycle [ 11 ] which optimizes the operation of the supervised elements and enables the realization of self- properties. Such a control loop continuously monitors and analyzes the system and its environment. Based on this information it plans the next steps and executes the planned actions. The different phases have access to a common knowledge base which provides information about the supervised elements or system.
Especially for automotive systems with various requirements and constraints, enabling self-adaptation and building a control architecture are a challenging task. Different aspects 4 star hotel barcelona of the automotive system like safety issues must be considered by the control components appropriately. Furthermore, 4 star hotel barcelona the control architecture has to react quickly to changing conditions, either 4 star hotel barcelona from inside the system (e.g., hardware or software failures) or from outside 4 star hotel barcelona the system (e.g., changing environmental conditions).
The control component of a self-adaptive system can be realized in different ways. Either a single-centralized control entity may realize the adaptation 4 star hotel barcelona of a software system, or multiple control components may realize the adaptation of composite of software systems in a decentralized manner [ 12 ]. While centralized 4 star hotel barcelona control may not be efficient for realizing adaptation in large, 4 star hotel barcelona complex, and heterogeneous systems (e.g., automotive embedded systems), fully decentralized approaches may lead to a coordination overhead within the system.
An alternative is hierarchical multi-layered control architectures, as discussed in [ 11 ], where multiple control loops cooperate to achieve 4 star hotel barcelona adaptation. In this work, we model such automotive systems using a set of constraints as introduced in [ 13 ]. Based on this, we can define the operation and responsibility of each local control cycle in the hierarchy based on different functional criteria. Thus, resulting in a control architecture 4 star hotel barcelona with a certain number of control loops arranged in a specific way using multiple layers. Even though all control cycles work on the same set of constrains, local responsibility 4 star hotel barcelona means that only some variables can be controlled and the others are fixed. This also means that some constraints are out of scope for some local control cycles. We compare different options, regarding different splits of responsibility as well as multi-layered control versus centralized control.
In summary, the main contributions of this paper are as follows. (i) We introduce 4 star hotel barcelona different instances of the multi-layered control approach for automotive embedded systems which hierarchically enforce system requirements on several layers. Therefore, the software components of the system are clustered based on different functional criteria realizing a control architecture with different numbers of layers. (ii) We compare these concrete instances of the multi-layered approach in a self-healing scenario with realistic setups of up to 100 ECUs. It is shown that local repair based on a layered control approach in such a system is more efficient than a pure central approach. Secondly, it shows that a responsibility split based on locality w.r.t. network topology performs better than a split regarding functional areas.
The structure of this paper is as follows. In Section 2 a brief introduction to automotive 4 star hotel barcelona embedded systems is given. Section 3 describes our approach for hierarchical, multi-layered control. Afterwards, we propose concrete 4 star hotel barcelona instances of our multi-layered control approach for self-adaptive automotive software system. In Section 5 , we outline how to realize 4 star hotel barcelona self-adaptation during runtime using our multi-layered 4 star hotel barcelona control approach. Moreover, we introduce a coordination mechanism 4 star hotel barcelona for the different control loops within 4 star hotel barcelona the multi-layered control architecture. Section 6 presents the results 4 star hotel barcelona of the experiments which we performed to compare our concrete instances of the multi-layered approach with a centralized control approach. Section 7 discusses related work.
An automotive embedded system is a distributed real-time system with heterogeneous hardware platforms (ECUs) interconnected by different network buses [ 5 ]. Moreover, the automotive embedded system consists of various functionalities which are implemented in software and which must satisfy 4 star hotel barcelona different requirements.
Typical constraints in terms of automotive systems concern hardware platform resources (e.g., volatile, non-volatile 4 star hotel barcelona memory), network resources (e.g., bandwidth), task dependencies, 4 star hotel barcelona task schedulability, timing, or the network topology. A detailed description of these equations, 4 star hotel barcelona called system 4 star hotel barcelona constraints, which define valid system configurations under real-time constraints, can be found in [ 13 ]. This set of equations is optimized to be solved efficiently during runtime in order to enable the computation of valid system configurations in an adaptive system in reasonable time. 3. Hierarchical, Multi-Layered Control Approach
To cope with the complexity of modern automotive embedded systems, we propose 4 star hotel barcelona a hierarchical, multi-layered control approach 4 star hotel barcelona (see Figure 1 ). Resources 4 star hotel barcelona for enforcing adaptations at runtime are scarce 4 star hotel barcelona but absolutely inevitable for realizing self- properties. Therefore, a divide-and-conquer strategy can be applied during design which partitions the automotive embedded system into smaller entities so-called clusters. Figure 1: Hierarchical, multi-layered control approach.
Repeated partitioning of the automotive embedded system results in a hierarchy of clusters, representing the entire system. Thereby, each cluster has at least one parent cluster and any number of child clusters. The top level of this hierarchy consists of exactly one cluster. Each cluster within the hierarchy is controlled by its own control loop, thus building a multi-layered control architecture. A control 4 star hotel barcelona loop is an external component which is not included within the cluster itself. It is supervising and controlling (adapting) the clustered elements during runtime, so that the constraints
Due to the repeated segmentation of the automotive embedded system, the clusters on the lowest layers in this control architecture are relatively small. 4 star hotel barcelona They consist of only a few software components. Due to the individual implementation of the control loops at the lowest layers, this hierarchical, multi-layered control approach can be tailored individually for the needs of specific functionalities or sub-systems of the automotive embedded system (e.g., safety-critical X-by-wire systems). As a drawback, the clusters 4 star hotel barcelona on the lower layers have a restricted local scope and are not always capable of determining a new, valid cluster configuration which satisfies all given requirements.
At the higher layers, the number of software components included in a cluster, which must be controlled, is growing. Thus, the possibilities of determining a valid cluster configuration increase but also th