Self-management of Ambient Intelligence Systems: a Pure Agent-based Approach ∗ (Extended Abstract) Inmaculada Ayala Universidad de Málaga Bulevar Louis Pasteur, 35 Málaga, Spain ayala@lcc.uma.es Mercedes Amor Universidad de Málaga Bulevar Louis Pasteur, 35 Málaga, Spain pinilla@lcc.uma.es Lidia Fuentes Universidad de Málaga Bulevar Louis Pasteur, 35 Málaga, Spain lff@lcc.uma.es ABSTRACT Ambient Intelligence systems (AmI) are normally composed of networked heterogeneous devices with critical resource limitations. One of the biggest requirements of AmI systems is that they should be capable of self-management in order to adapt their behavior and resources to environmental con- ditions and variable device resources. Autonomous agents are a good option to endow AmI systems with self-managing capabilities, but current agent platform implementations do not adequately address the heterogeneity requirements of AmI systems, given the impossibility until now of producing pure agent-based solutions. In this paper we present a pure agent-based solution for self-managing AmI systems, with particular emphasis on defining a working solution consid- ering the diversity of devices and communication protocols through which AmI devices must interoperate. Categories and Subject Descriptors I.2.11 [Computing Methodologies]: Distributed Artifi- cial Intelligence—Multiagent systems General Terms Design, Experimentation, Management Keywords Agent Oriented Software Engineering, Self-management, Am- bient Intelligence, Lightweight devices 1. INTRODUCTION Ambient Intelligence (AmI) environments represent a new generation of computing systems equipped with devices with special capabilities that make people aware of the environ- ment and react to it, in a more natural way [4]. These systems are composed of a large variety of networked hetero- geneous devices, such as mobile phones or Wireless Sensors ∗ This work has been supported by the Project RAP TIN2008-01942 and the Project FamWare P09-TIC-5231. Appears in: Proceedings of the 11th International Con- ference on Autonomous Agents and Multiagent Systems (AAMAS 2012), Conitzer, Winikoff, Padgham, and van der Hoek (eds.), 4-8 June 2012, Valencia, Spain. Copyright c ⃝ 2012, International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org). All rights reserved. Networks (WSNs). Normally, the majority of AmI devices show symptoms of degradation, such as energy loss or failure of some network nodes, which requires explicit management action, for example saving energy to guarantee the system’s survival. Consequently AmI systems demand the reconfig- uration of their internal functioning in response to changes in their environment. This means that AmI systems must behave as autonomic systems with a self-managing capacity. The self-management properties are inspired by the prop- erties of agents [5], such as autonomy, distribution and proac- tiveness. This leads us to consider agents and Multi-Agent Systems (MASs) as effective metaphors for system design and implementation of AmI scenarios. In this paper we will focus on how an agent-based solution can help to im- plement the self-management requirement of AmI systems. Several approaches already exploit MAS in the context of self-managing AmI systems [2][3], but the solution proposed by most of them cannot be considered a pure agent-based solution since agents are just used to apply Artificial Intel- ligence techniques (e.g. learning or planning algorithms), or as autonomic managers of an add-on autonomic system.These solutions are not energy efficient, as the self-management tasks imply an extra traffic between the autonomic agent manager(s) and the managed devices. So, pure agent-based solutions are more energy efficient, since they minimize the self-management traffic, which can be considerable in AmI systems with for example thousands of sensors. In this paper we introduce self-StarMAS, a set of coop- erating agents, running in each device of an AmI system, able to communicate and interoperate through heteroge- nous communication protocols, and with the capacity of self- management adapted to each kind of agent/device. 2. THE SELF-STARMAS SYSTEM In this section we will present the self-StarMAS system (see fig. 1) focusing on the challenges that pose the use of agent technology to self-manage AmI systems and how they are addressed by our approach. The use of agents to support self-management is not a new research area, but we consider that current solutions do not address all these challenges well. C1 - Self-management: Much research on self-mana- gement is progressing and several self-* properties intro- duced by IBM can be considered a good starting point (self- configuring, self-optimizing,...). Achieving: We propose a MAS composed by a set of cooperating agents with the