Embedded Automation in Human-Agent Environment

This book has resulted from ongoing research by a group of enthusiastic people within both Defence Science and Technology Organisation (DSTO) and Knowledge- Based Intelligent Information and Engineering Systems (KES) Centre during work on automation and Artificial Intelligence (AI). Our effort has enabled the formation of teams that combine the skills of both human and machine (electronic) members. This book proposes a general, conceptual framework for the development of automation in human-machine environments that will allow such teams to work effectively and efficiently. This chapter introduces the subject and the main challenges which developers have to face. The method of approach in developing this conceptual framework will be described with an outline of original contributions. An overview of the structure of the book and its chapters is also presented.

This chapter defines the related terms for this topic and key issues surrounding the evolution of the science supporting the CI domain, together with an introduction of several of the tools and training practices developed to support the research in this area.World war two introduced many new technologies and expanded the engineering domain so rapidly that any impediments in a specific research topics were being abandoned in favor of more productive exploits. Over half a century has passed and we are still using the same fundamental computing architectures. Many disciplines have contributed to the development of agents, threads and component architecture. This chapter briefly discusses many of the key developments in CI and their relationship to this book.

This chapter presents a brief introduction to existing research in Knowledge-Based engineering to achieve automation using IAs. It defines how the term intelligence is determined and the tools used to exploit human machine interaction. So what is Intelligence, why use AI and which architecture provides the best choice of tools. A whole field of science has developed around AI and is based predominantly on computer technology and the enhancements used to develop their capability. The literature illustrates the evolution of the key sciences used to support AI and distinguishes between human, machine and the architectures required to solve problems. Disruptions in developing AI techniques bounded with insertion of new technology, but most were focused on commercial applications, other than niche domains like AI.

This chapter briefly discusses how technology evolved during the computer evolution, followed by types of agent architectures that have surfaced and how they are being used in modern systems. A brief description concludes with a discussion of future trends.

In this chapter a brief description of syntax, semantics, topology, ontology and taxonomy is discussed in relation to agency theory. The Open Systems Interconnection (OSI) model is included to help aid the ready align the technical, physical and cognitive connections. At present the communications layer is being embedded into many agent architectures, when in fact, it should remain as an underlying transport mechanism. However the interoperable functions are still required as are a standard methodology of exchanging information and intent. These and related subjects mechanisms are also included to generate further discussion into what the future standard should include.

The question of autonomy using agents is complex and requires significant skill to write applications that interoperate successfully. The complexity is derived from the need to cater for every conceivable outcome. Providing applications with the ability of being able to adapt dynamically is vital in this domain.Abrief discussion of design patterns, threads, components and associated architectural research are used to assist the reader in understanding this topic.

Agent technologies have been an ongoing field of research that has failed to produce a solution for a complex, dynamic, real-world problem to the scale and expectations the technology is perceived to deliver.Web based-applications represent shop fronts and transactional services. However due to security, response times and reliability, industry has chosen to avoid them for mainstream advanced information process technologies. The KES centre has conducted reviews on agents and agent based systems and concluded that industry requires a BDI framework within an interoperable team-based environment. After reviewing projects using R-CAST and JACK like Integration of Reactive Behavior and Rational Planning (InteRRaP) [260], alGOl Logical prOGramming (Golog) [222] and RETSINA [145], the risk of mode confusion by human operators surfaced as a critical issue. In many cases, load sharing can alleviate this problem, hence the requirement for agents interoperability, especially between supervisors, siblings and subordinates. Trust and negotiation become important factors in establishing context, cognition and task related activities. Such attributes are normally coded or hidden within an applications corporate processes. Autonomous teams with the ability to dynamically alter functionality present ideal enhancements to any scenario. Modularised capabilities must be, proven available before they are adopted by industry. to achieve reusable, hardened applications.

An agentmay be described as anything physical, synthetic or coded that is perceived of being capable of interacting upon an environment. A human agent would be seen to have sensors (eyes, ears, and other organs) to create percepts of the environment and effectors (hands, legs, mouth, and other body parts) to act upon the environment. A robotic agent may substitute cameras or other sensors to perceive the current situation, while various mechanised attachments could be used to effect some action within that environment. When defining an agent, researchers describe the properties it should exhibit. The first property is autonomy,which means operating without the direct intervention of humans. Second is social ability which describes the ability to interact with other agents, agent applications and/or even humans. Third is reactivity, which includes a means of perceiving the environment and responding to any changes that occur within it at a given point in time. Finally, pro-activeness means exhibiting goal-directed behavior [407]. There are many architectures domains of influences and technologies that embody agent systems. When implemented as a system, agents are capable of achieving highly sophisticated goals autonomously and if written correctly,will continue to find a solution until the goal is complete. This chapter investigates how agents architectures evolved, the level of control, construction and mobility. Discussion continues to explore communications, how data is passed or concepts merged and the technologies used.

In this chapter we describe the requirements needed to construct an Agent Factory. We discuss how many of the fundamental Java capabilities can now be used to build credible applications to solve enterprise style problems.Abrief list of recent enhancements are provided with a number of future changes flagged.

Many algorithms exist to solve all sorts of puzzle. A lot of these have been implemented in software. The reader is invited to seek out alternative solutions and are challenged to provide their own implementation(s). Russell and Norvig [320] provide the necessary background to the history and adaptation of many of these techniques within the AI domain. After introducing the problem, we will briefly discuss Heuristic search [16], Brute Force (exhaustive) and Backtracking [401]. The only mandatory condition for a Sudoku puzzle is that it must have a solution, although it should be solved logically and without guessing or search.

In this chapter we work through a number of techniques raised in previous chapters. Each of the examples is accompanied with a brief explanation, although the syntax within some listings have been abbreviated to simplify the description.We introduced both static and dynamic methods associated with deriving results. A number of algorithms are then explained before discussing the MVC pattern associated with the Sudoku solver application.