Load–frequency control : a GA-based multi-agent reinforcement learning

The load–frequency control (LFC) problem has been one of the major subjects in a power system. In practice, LFC systems use proportional–integral (PI) controllers. However since these controllers are designed using a linear model, the non-linearities of the system are not accounted for and they are incapable of gaining good dynamical performance for a wide range of operating conditions in a multi-area power system. A strategy for solving this problem because of the distributed nature of a multi-area power system is presented by using a multi-agent reinforcement learning (MARL) approach. It consists of two agents in each power area; the estimator agent provides the area control error (ACE) signal based on the frequency bias estimation and the controller agent uses reinforcement learning to control the power system in which genetic algorithm optimisation is used to tune its parameters. This method does not depend on any knowledge of the system and it admits considerable flexibility in defining the control objective. Also, by finding the ACE signal based on the frequency bias estimation the LFC performance is improved and by using the MARL parallel, computation is realised, leading to a high degree of scalability. Here, to illustrate the accuracy of the proposed approach, a three-area power system example is given with two scenarios.

Modelling issues on the railway resource management process using multi-agent system

Many infrastructure and necessity systems such as electricity and telecommunication in Europe and the Northern America were used to be operated as monopolies, if not state-owned. However, they have now been disintegrated into a group of smaller companies managed by different stakeholders. Railways are no exceptions. Since the early 1980s, there have been reforms in the shape of restructuring of the national railways in different parts of the world. Continuous refinements are still conducted to allow better utilisation of railway resources and quality of service. There has been a growing interest for the industry to understand the impacts of these reforms on the operation efficiency and constraints. A number of post-evaluations have been conducted by analysing the performance of the stakeholders on their profits (Crompton and Jupe 2003), quality of train service (Shaw 2001) and engineering operations (Watson 2001). Results from these studies are valuable for future improvement in the system, followed by a new cycle of post-evaluations. However, direct implementation of these changes is often costly and the consequences take a long period of time (e.g. years) to surface. With the advance of fast computing technologies, computer simulation is a cost-effective means to evaluate a hypothetical change in a system prior to actual implementation. For example, simulation suites have been developed to study a variety of traffic control strategies according to sophisticated models of train dynamics, traction and power systems (Goodman, Siu and Ho 1998, Ho and Yeung 2001). Unfortunately, under the restructured railway environment, it is by no means easy to model the complex behaviour of the stakeholders and the interactions between them. Multi-agent system (MAS) is a recently developed modelling technique which may be useful in assisting the railway industry to conduct simulations on the restructured railway system. In MAS, a real-world entity is modelled as a software agent that is autonomous, reactive to changes, able to initiate proactive actions and social communicative acts. It has been applied in the areas of supply-chain management processes (García-Flores, Wang and Goltz 2000, Jennings et al. 2000a, b) and e-commerce activities (Au, Ngai and Parameswaran 2003, Liu and You 2003), in which the objectives and behaviour of the buyers and sellers are captured by software agents. It is therefore beneficial to investigate the suitability or feasibility of applying agent modelling in railways and the extent to which it might help in developing better resource management strategies. This paper sets out to examine the benefits of using MAS to model the resource management process in railways. Section 2 first describes the business environment after the railway 2 Modelling issues on the railway resource management process using MAS reforms. Then the problems emerge from the restructuring process are identified in section 3. Section 4 describes the realisation of a MAS for railway resource management under the restructured scheme and the feasible studies expected from the model.

Characterizing multi-agent team behavior from partial team tracings : evidence from the English Premier League

Real-world AI systems have been recently deployed which can automatically analyze the plan and tactics of tennis players. As the game-state is updated regularly at short intervals (i.e. point-level), a library of successful and unsuccessful plans of a player can be learnt over time. Given the relative strengths and weaknesses of a player’s plans, a set of proven plans or tactics from the library that characterize a player can be identified. For low-scoring, continuous team sports like soccer, such analysis for multi-agent teams does not exist as the game is not segmented into “discretized” plays (i.e. plans), making it difficult to obtain a library that characterizes a team’s behavior. Additionally, as player tracking data is costly and difficult to obtain, we only have partial team tracings in the form of ball actions which makes this problem even more difficult. In this paper, we propose a method to overcome these issues by representing team behavior via play-segments, which are spatio-temporal descriptions of ball movement over fixed windows of time. Using these representations we can characterize team behavior from entropy maps, which give a measure of predictability of team behaviors across the field. We show the efficacy and applicability of our method on the 2010-2011 English Premier League soccer data.

Attitudes, Ideologies and Self-Organisation : Information Load Minimisation in Multi-Agent Decision Making

Sophisticated models of human social behaviour are fast becoming highly desirable in an increasingly complex and interrelated world. Here, we propose that rather than taking established theories from the physical sciences and naively mapping them into the social world, the advanced concepts and theories of social psychology should be taken as a starting point, and used to develop a new modelling methodology. In order to illustrate how such an approach might be carried out, we attempt to model the low elaboration attitude changes of a society of agents in an evolving social context. We propose a geometric model of an agent in context, where individual agent attitudes are seen to self-organise to form ideologies, which then serve to guide further agent-based attitude changes. A computational implementation of the model is shown to exhibit a number of interesting phenomena, including a tendency for a measure of the entropy in the system to decrease, and a potential for externally guiding a population of agents towards a new desired ideology.

Reliable, cloud-based communication for multi-robot systems

In contrast to single robotic agent, multi-robot systems are highly dependent on reliable communication. Robots have to synchronize tasks or to share poses and sensor readings with other agents, especially for co-operative mapping task where local sensor readings are incorporated into a global map. The drawback of existing communication frameworks is that most are based on a central component which has to be constantly within reach. Additionally, they do not prevent data loss between robots if a failure occurs in the communication link. During a distributed mapping task, loss of data is critical because it will corrupt the global map. In this work, we propose a cloud-based publish/subscribe mechanism which enables reliable communication between agents during a cooperative mission using the Data Distribution Service (DDS) as a transport layer. The usability of our approach is verified by several experiments taking into account complete temporary communication loss.

An efficient and light-weight multi-tasking model to achieve Intra-Agent concurrency for SAGE

Multi-agent systems implicate a high degree of concurrency at both the Inter- and Intra-Agent levels. Scalable, fault tolerant, Agent Grooming Environment (SAGE), the second generation, FIPA compliant MAS requires a built in mechanism to achieve both the Inter- and Intra-Agent concurrency. This paper dilates upon an attempt to provide a reliable, efficient and light-weight solution to provide intra-agent concurrency with-in the internal agent architecture of SAGE. It addresses the issues related to using the JAVA threading model to provide this level of concurrency to the agent and provides an alternative approach that is based on an eventdriven, concurrent and user-scalable multi-tasking model for the agent's internal model. The findings of this paper show that our proposed approach is suitable for providing an efficient and lightweight concurrent task model for SA GE and considerably outweighs the performance of multithreaded tasking model based on JAVA in terms of throughput and efficiency. This has been illustrated using the practical implementation and evaluation of both models. © 2004 IEEE.

Multi-agent search using?sensors with heterogeneous capabilities

In this thesis we address the problem of multi-agent search. We formulate two deploy and search strategies based on optimal deployment of agents in search space so as to maximize the search effectiveness in a single step. We show that a variation of centroidal Voronoi configuration is the optimal deployment. When the agents have sensors with different capabilities, the problem will be heterogeneous in nature. We introduce a new concept namely, generalized Voronoi partition in order to formulate and solve the heterogeneous multi-agent search problem. We address a few theoretical issues such as optimality of deployment, convergence and spatial distributedness of the control law and the search strategies. Simulation experiments are carried out to compare performances of the proposed strategies with a few simple search strategies.

Multi-agent Search using Voronoi partitions

This paper addresses the problem of multiagent search in an unknown environment. The agents are autonomous in nature and are equipped with necessary sensors to carry out the search operation. The uncertainty, or lack of information about the search area is known a priori as a probability density function. The agents are deployed in an optimal way so as to maximize the one step uncertainty reduction. The agents continue to deploy themselves and reduce uncertainty till the uncertainty density is reduced over the search space below a minimum acceptable level. It has been shown, using LaSalle’s invariance principle, that a distributed control law which moves each of the agents towards the centroid of its Voronoi partition, modified by the sensor range leads to single step optimal deployment. This principle is now used to devise search trajectories for the agents. The simulations were carried out in 2D space with saturation on speeds of the agents. The results show that the control strategy per step indeed moves the agents to the respective centroid and the algorithm reduces the uncertainty distribution to the required level within a few steps.

Multi-agent Reinforcement Learning for Traffic Signal Control

Optimal control of traffic lights at junctions or traffic signal control (TSC) is essential for reducing the average delay experienced by the road users amidst the rapid increase in the usage of vehicles. In this paper, we formulate the TSC problem as a discounted cost Markov decision process (MDP) and apply multi-agent reinforcement learning (MARL) algorithms to obtain dynamic TSC policies. We model each traffic signal junction as an independent agent. An agent decides the signal duration of its phases in a round-robin (RR) manner using multi-agent Q-learning with either is an element of-greedy or UCB 3] based exploration strategies. It updates its Q-factors based on the cost feedback signal received from its neighbouring agents. This feedback signal can be easily constructed and is shown to be effective in minimizing the average delay of the vehicles in the network. We show through simulations over VISSIM that our algorithms perform significantly better than both the standard fixed signal timing (FST) algorithm and the saturation balancing (SAT) algorithm 15] over two real road networks.

Multi-agent modelling for the simulation of a simple smart microgrid

The smart grid is a highly complex system that is being formed from the traditional power grid, adding new and sophisticated communication and control devices. This will enable integrating new elements for distributed power generation and also achieving an increasingly automated operation so for actions of the utilities as for customers. In order to model such systems a bottom-up method is followed, using only a few basic elements which are structured into two layers: a physical layer for the electrical power transmission, and one logical layer for element communication. A simple case study is presented to analyse the possibilities of simulation. It shows a microgrid model with dynamic load management and an integrated approach that can process both electrical and communication flows.

Nash equilibria in multi-agent motor interactions.

Social interactions in classic cognitive games like the ultimatum game or the prisoner's dilemma typically lead to Nash equilibria when multiple competitive decision makers with perfect knowledge select optimal strategies. However, in evolutionary game theory it has been shown that Nash equilibria can also arise as attractors in dynamical systems that can describe, for example, the population dynamics of microorganisms. Similar to such evolutionary dynamics, we find that Nash equilibria arise naturally in motor interactions in which players vie for control and try to minimize effort. When confronted with sensorimotor interaction tasks that correspond to the classical prisoner's dilemma and the rope-pulling game, two-player motor interactions led predominantly to Nash solutions. In contrast, when a single player took both roles, playing the sensorimotor game bimanually, cooperative solutions were found. Our methodology opens up a new avenue for the study of human motor interactions within a game theoretic framework, suggesting that the coupling of motor systems can lead to game theoretic solutions.

Adaptive uncertain information fusion to enhance plan selection in BDI agent systems

Correctly modelling and reasoning with uncertain information from heterogeneous sources in large-scale systems is critical when the reliability is unknown and we still want to derive adequate conclusions. To this end, context-dependent merging strategies have been proposed in the literature. In this paper we investigate how one such context-dependent merging strategy (originally defined for possibility theory), called largely partially maximal consistent subsets (LPMCS), can be adapted to Dempster-Shafer (DS) theory. We identify those measures for the degree of uncertainty and internal conflict that are available in DS theory and show how they can be used for guiding LPMCS merging. A simplified real-world power distribution scenario illustrates our framework. We also briefly discuss how our approach can be incorporated into a multi-agent programming language, thus leading to better plan selection and decision making.

An Architecture of a Multi-Agent System for SCADA - Dealing With Uncertainty, Plans and Actions

This paper presents a multi-agent system approach to address the difficulties encountered in traditional SCADA systems deployed in critical environments such as electrical power generation, transmission and distribution. The approach models uncertainty and combines multiple sources of uncertain information to deliver robust plan selection. We examine the approach in the context of a simplified power supply/demand scenario using a residential grid connected solar system and consider the challenges of modelling and reasoning with
uncertain sensor information in this environment. We discuss examples of plans and actions required for sensing, establish and discuss the effect of uncertainty on such systems and investigate different uncertainty theories and how they can fuse uncertain information from multiple sources for effective decision making in
such a complex system.