Urban Richter

Adaptivity and self-organization in organic computing systems

Organic Computing (OC) and other research initiatives like Autonomic Computing or Proactive Computing have developed the vision of systems possessing life-like properties: they self-organize, adapt to their dynamically changing environments, and establish other so-called self-x properties, like self-healing, self-configuration, self-optimization, etc. What we are searching for in OC are methodologies and concepts for systems that allow to cope with increasingly complex networked application systems by introduction of self-x properties and at the same time guarantee a trustworthy and adaptive response to externally provided system objectives and control actions. Therefore, in OC, we talk about controlled self-organization. Although the terms self-organization and adaptivity have been discussed for years, we miss a clear definition of self-organization in most publications, which have a technically motivated background. In this article, we briefly summarize the state of the art and suggest a characterization of (controlled) self-organization and adaptivity that is motivated by the main objectives of the OC initiative. We present a system classification of robust, adaptable, and adaptive systems and define a degree of autonomy to be able to quantify how autonomously a system is working. The degree of autonomy distinguishes and measures external control that is exerted directly by the user (no autonomy) from internal control of a system which might be fully controlled by an observer/controller architecture that is part of the system (full autonomy). The quantitative degree of autonomy provides the basis for characterizing the notion of controlled self-organization. Furthermore, we discuss several alternatives for the design of organic systems.

Observation and Control of Organic Systems

Organic Computing (OC) assumes that current trends and recent developments in computing, like growing interconnectedness and increasing computational power, pose new challenges to designers and users. In order to tackle the upcoming demands, OC has the vision to make systems more life-like (organic) by endowing them with abilities such as self-organisation, self-configuration, self-repair, or adaptation. Distributing computational intelligence by introducing concepts like self-organisation relieves the designer from exactly specifying the low-level system behaviour in all possible situations. In addition, the user has the possibility to define a few high-level goals, rather than having to manipulate many low-level parameters.

Decentralized energy-management to control smart-home architectures

In this paper, we focus on a real world scenario of managing electrical demand sets of a smart-home. External signals, reflecting the low voltage grids state, are used to support the challenge of balancing energy demand and generation. To manage the smart-homes appliances and to integrate electric vehicles as energy storages decentralized control systems are investigated.

Towards a quantitative notion of self-organisation

Organic computing (OC) and other research initiatives like autonomic computing or proactive computing have developed the idea of systems that possess life-like properties, that self-organise, that adapt to their dynamically changing environments, and that establish other so-called self-x properties, like self-healing, self-configuration, self-optimisation etc. What we are searching for in OC are not concepts for systems that simply self-organise, but systems that self-organise to achieve a well defined system goal. Therefore we talk in OC about controlled self-organisation. Although the term self-organisation has been discussed for years, we miss a clear definition of self-organisation in most publications, which have a technically motivated background. In this paper, we summarise the state of the art and introduce a definition of self-organisation that addresses the problem of designing self-organising technical systems, which is the main objective of the OC initiative.

Measurement and control of self-organised behaviour in robot swarms

Todays technical systems are becoming increasingly complex. Future systems will consist of a multitude of complex soft- and hardware components, which interact with each other to satisfy global system functional requirements. This trend bears the risk of more and more breakdowns and other unexpected behaviour. Organic Computing (OC) has the vision of addressing the challenges of complex distributed systems by making them more life-like (organic), i. e. endowing them with abilities such as self-organisation, self-configuration, self-repair, or adaptation. This can only be achieved by giving the system elements adequate degrees of freedom. This may result in an emergent behaviour, which can be positive as well as negative. Therefore, we need an observer/ controller architecture, which allows for self-organisation but at the same time enables adequate reactions to control the - sometimes completely unexpected - emerging global behaviour. In this paper, we give an introduction to a generic observer/controller architecture, adapt this framework to a scenario of a self-organising robot swarm, and show how to control and prevent global, collective, unwanted behaviour based on observations of the local behaviour of the distributed agents.

Improving XCS Performance by Distribution

Learning Classifier Systems (LCSs) are rule-based evolutionary reinforcement learning (RL) systems. Today, especially variants of Wilsons eXtended Classifier System (XCS) are widely applied for machine learning. Despite their widespread application, LCSs have drawbacks: The number of reinforcement cycles an LCS requires for learning largely depends on the complexity of the learning task. A straightforward way to reduce this complexity is to split the task into smaller sub-problems. Whenever this can be done, the performance should be improved significantly. In this paper, a nature-inspired multi-agent scenario is used to evaluate and compare different distributed LCS variants. Results show that improvements in learning speed can be achieved by cleverly dividing a problem into smaller learning sub-problems.

Using organic computing to control bunching effects

In this paper, the well-known emergent phenomenon of bunching as appearing in lift group traffic control systems is taken as a technical scenario for validating the generic observer/controller architecture which has been designed as part of an anticipated organic framework - providing generic toolbox mechanisms to observe, analyse, and control emergent behaviour in self-organising systems. In particular, we show how to control and prevent global, collective, unwanted behaviour of groups of lifts, based on observations of the local behaviour of lift cabins.

Adaption of XCS to multi-learner predator/prey scenarios

Learning classifier systems (LCSs) are rule-based evolutionary reinforcement learning systems. Today, especially variants of Wilsons extended classifier system (XCS) are widely applied for machine learning. Despite their widespread application, LCSs have drawbacks, e. g., in multi-learner scennarios, since the Markov property is not fulfilled. In this paper, LCSs are investigated in an instance of the generic homogeneous and non-communicating predator/prey scenario. A group of predators collaboratively observe a (randomly) moving prey as long as possible, where each predator is equipped with a single, independent XCS. Results show that improvements in learning are achieved by cleverly adapting a multi-step approach to the characteristics of the investigated scenario. Firstly, the environmental reward function is expanded to include sensory information. Secondly, the learners are equipped with a memory to store and analyze the history of local actions and given payoffs.

Collaborating and Learning Predators on a Pursuit Scenario

A generic predator/prey pursuit scenario is used to validate a common learning approach using Wilson’s eXtended Learning Classifier System (XCS). The predators, having only local information, should independently learn and act while at the same time they are urged to collaborate and to capture the prey. Since learning from scratch is often a time consuming process, the common learning approach, as investigated here, is compared to an individual learning approach of selfish learning agents. A special focus is set on the performance of how quickly the team goal is achieved in both learning scenarios. This paper provides new insights of how agents with local information could learn collaboratively in a dynamically changing multi-agent environment. Furthermore, the concept of a common rule base based on Wilson’s XCS is investigated. The results based on the common rule base approach show a significant speed up in the learning performance but may be significantly inferior on the long run, in particular in situations with a moving prey.