The field of Artificial Intelligence (AI) has increasingly expanded to create intelligent behaviour and consequently has improved the system(s) efficiency and performance. Automation had become necessary in various applications where human intervention necessarily needed to be minimised. In this research, the Multi-Agent System (MAS) was developed to improve autonomous navigation in a dynamic environment through agent teaming and learning. The MAS consists of Intelligent Agents (IAs), which are autonomous computational entities capable of perceiving and reacting to environmental changes; communicating and interacting with each other; and bringing a series of agent-instigated goal actions to fruition. AIs are equipped with unique capabilities and thus share knowledge and skills through Agent Communication Languages (ACLs) to achieve a common goal(s).
A key concept of this research was the hybridisation of two well-established AI techniques through the sharing of knowledge and capabilities in Multi-Agent Systems. These techniques include the path-finding algorithm of A* and the machine-learning algorithm of real-time Neuro-Evolution of Augmenting Topologies (rtNEAT). Each of these techniques is suitable for specific domains. The A* algorithm is a graph search algorithm, which is capable of finding the most efficient path between two points in a digital world using navigation graphs. The rtNEAT algorithm is an Evolutionary Artificial Neural Network (EANN) algorithm used for evolving agent behaviour through optimising the structure of the Artificial Neural Network (ANN) that controls its actions.
The Multi-Agent System developed in this research joined with the isolated subfields of Artificial Intelligence and through this hybridisation, the navigation of Intelligent Agents in a stochastic and dynamic environment was much improved. Agents within Multi-Agent System were given different roles based on individual capabilities. Some agents acted as supervisors to facilitate the interaction and communication between other agents and also managed tasks by assigning them to agents with required capabilities. A category of agents was then equipped with sensors and evolved using rtNEAT to optimise their sensors. As a consequence of learning, they were able to avoid collisions while exploring the unknown virtual world. Furthermore, as they discovered the environment, they engaged in knowledge sharing to create a digitised representation. This representation was then used by another category of agents equipped with path-finding capabilities to navigate efficiently and collect resources cooperatively.
A concept demonstrator was developed as part of this research to investigate the interactions and communication between agents and demonstrated these results as part of a resource-gathering scenario in a virtual world. A detailed discussion of Artificial Intelligence (AI), Intelligent Agent (IA), Multi-Agent System (MAS), Agent Communication Language (ACL), Agent Learning, and Path-finding algorithms occur throughout the research.
Overview of my Research on AI, Machine Learning and Multi-Agent Systems:
The following is the list of topics I covered in my research on Artificial Intelligence, Machine Learning and Multi-Agent Systems. I provide a brief summary of each section and more details are provided in the link to the articles on this website.
Motivation for Research on Artificial Intelligence
My Motivation for Research on Artificial Intelligence: AI techniques have significant potential applications in many fields. AI research aims to create autonomous entities that are capable of acting reliably on behalf of humans. Such systems would have great advantages in domains where human intervention needs to be minimised. This section discusses my motivation for researching Artificial Intelligence, Machine Learning and Multi-Agent Systems.
A Brief History of Artificial Intelligence
A Brief History of Artificial Intelligence provides a short history of Artificial Intelligence from traditional techniques of rule-based AI, FSMs and Fuzzy Logic to more advanced AI techniques used for learning, and looks at evolving autonomous behaviours in stochastic and dynamic environments.
Some Applications of Artificial Intelligence
AI techniques are used to tackle difficult tasks and solve problems from small systems to complex mission-critical systems. Automation is convenient and is becoming commonly used for dangerous or repetitive tasks. Some examples of the applications of AI are discussed in this section.
Testbeds for Artificial intelligence Research
Computer games and simulations have contributed much to the research and development of AI techniques in a variety of applications. These tools have provided testing environments for the implementation and testing of AI theories and the creation of intelligent systems. Some examples of these Testbeds for Artificial intelligence Research are discussed in this article.
What is Artificial Intelligent Agent?
This article begins with a discussion of different definitions proposed for describing agents and Intelligent Agents, and the characteristics of an Intelligent Agent, which include autonomy, reactivity, proactivity, social ability, mobility, learning, and rationality.
What is a Multi-Agent System?
A Multi-Agent System is a system that consists of multiple agents that communicate and interact with each other to achieve a certain goal(s). In this article, we will go through what Multi-Agent Systems are and describe their characteristics.
Knowledge Representation in Artificial Intelligence
Knowledge has to be represented in a meaningful way for agents to be able to understand the information transmitted between them. This section discusses the concept of knowledge representation in Artificial Intelligence and mentioned some popular examples of languages used for defining ontology and the body of knowledge including Knowledge Interchange Format and Extensible Mark-up Language (XML).
Agent Communication Languages
Communication and knowledge-sharing characteristics are necessary in order to achieve interoperability in Multi-Agent System. This section discusses some of the Agent Communication Languages that have been developed to standardize the way agents communicate, including KQML, FIPA and SOAP.
Neural Network and Machine Learning Paradigms
Learning provides an agent with the ability to operate in unknown situations and environments. This article begins with a brief overview of learning techniques by discussing Artificial Neural Networks, their architecture, and different learning paradigms.
Evolutionary Algorithms: Genetic Algorithm and Neuro-Evolution Algorithms
This article covers the Evolutionary Algorithm techniques, in particular, the Genetic Algorithm and Evolutionary Artificial Neural Network techniques, which provide effective solutions by combining Neural Networks and Evolutionary Algorithms. An example of these techniques is the Neuroevolution of augmenting topologies (NEAT).
Neuro-Evolution of Augmenting Topologies Algorithm
This article explains the Neuro-Evolution of Augmenting Topologies (NEAT) and real-time Neuro-Evolution of Augmenting Topologies (rtNEAT) algorithms, discussing how they can endow agents with learning capabilities.
Agents in this research are equipped with sensors, which provide them with the ability to sense the unknown environment. The behavior of an agent is controlled by a neural network, which reads its onboard sensors and takes the appropriate actions to explore the environment. The learning ability allows agents to optimize their sensors and improve their actions of avoiding obstacles and navigating more quickly. The functionality of these sensors, along with the learning procedure implemented to optimize them, is discussed in this article.
Bibliography References for my Research on AI
And finally, this is the list of Bibliography References and the Glossary of Acronyms used in my research on Artificial Intelligence, Machine Learning and Multi-Agent Systems
Keywords: Artificial Intelligence, Multi-Agent Systems, Intelligent Agents, Real-time Neuro-Evolution of Augmenting Topologies, A* Path-finding Algorithm, Machine Learning, Agent Teaming, Agent Communication Languages, Dynamic Environment.
Pingback: Agent Communication Languages in AI for Knowledge Sharing Brainy Loop