Considering system design in work environments, engineers are interested to build adequate tools and support systems for human operators for the sake of good, satisfying work results. The engineers' focus lies on the issue of work system synthesis rather than system analysis, thereby considering the human operator as a given, sufficiently known system element, taking into account the advanced findings about the cognitive behaviour of humans. The knowledge about human cognition is of great benefit, too, to promote the right ideas in the course of the endeavours of computer science people and engineers to achieve artificial cognition with capabilities equivalent or even better than those of the human species. With the advances in computer science, engineers are more and more looking at designing systems which are relying heavily on artificial cognitive functions similar to those of humans. Researchers having a background in biological cybernetics, describes a fascinating series of experiments with toy vehicles, steadily being amended step by step until a kind of cognitive vehicle is achieved which behaves as it were controlled in a human-like fashion when moving around in its little toy world.
He claims that by introducing two particular types of electronic core elements simulating certain neural properties he knows from his neuroscience experiments with animal brains, networking them in certain structures and adding sensors and motors, human-like behaviour in vehicle guidance and control is possible. This includes functional capabilities of cognition such as knowledge storage capability, knowledge acquisition capability about objects and concepts pertinent to the vehicle environment (static and dynamic), including also the mapping of the environment, thereby achieving the capability of process prediction. Among the vehicles also those are outlined which can cope with more rare situations when the prediction does not correspond with what actually happens. Here, the additional function of a short-term memory is introduced in the toy vehicle design new solutions, for instance for the identification of new situations, yielding the capability of exploring knowledge already incorporated in the brain.
Eventually, in the last one of the succession of steadily upgraded vehicle designs, the capability of goal-directed behaviour based on explicitly stored desires is incorporated. It has already been perfectly illustrated by means of fascinatingly simple featuring of the crucial design steps how to accomplish the synthetic task of providing artificial cognition. His article is bridging thereby the disciplines of neuroscience, computer science, and engineering. It is strongly encouraging and stimulating for researchers of computer science and engineers to work for the realisation of artificial cognition in vehicles in real world applications. We can call the Braitenberg vehicles robots, if we look at them from the robotic community's point of view. We also can call them cognitive vehicles, if we look at them from the point of view of designing artificial cognition as such, which addresses not the whole vehicle system but just the part which is responsible of information processing in the vehicle by means of computers. Both perspectives can be subsumed under the theory of agents, then distinguishing between hardware agents (e.g. robots) and software agents (e.g. internet machines). Both kinds of agents can be part of work systems. Software agents need communication links to their environment within and outside the work system, whereas hardware agents also observe the environment in a more self-reliant fashion by means of sensing capabilities before acting on it. A typical example for a hardware agent might be a semi-autonomous unmanned vehicle. From the system engineer's point of view, agents in work systems are supposed to be a kind of artificial equivalent to human operators. In that sense, we follow the characterisation of agents as stated: Agents "must be
• Semi-autonomous: given a vague and imprecise specification, it must determine how the problem is best solved and then solve it, without constant guidance from the user.
• proactive: it should not wait to be told what to do next, rather it should make suggestions to the user
• responsive: it should take account of changing user needs and changes in the task environment
• adaptive: it should come to know user's preferences and tailor interactions to reflect these" In order to ensure these characteristics it has to be made use of capabilities like cognition, co-operation, communication, mobility, and learning at the extent needed.
It has to be noted though at this point that, by virtue, there are differences between the humans and artificial cognitive systems, mostly even huge differences. These differences are caused, in the first place, by the obvious differences regarding the physical implementation of the cognitive functions. As a consequence, depending on the technical implementation and the task requirements, the behavioural results of artificial cognitive systems might show both, better or worse performance than those of humans when encountering certain work situations. This has to be accounted for when designing artificial cognitive systems for use in work systems. Thus, agents are not necessarily designed as copies of humans. They rather are designed to do their job most effectively, regarding the function they have to carry out within a work system. There are applications, where the cognitive behaviour of an agent is deliberately designed to differ from that one of humans. At the other extreme there are applications, though, where the design goal is to achieve cognitive behaviour as similar as possible to humans. So far, agents are deployed in work systems as part of the operation-supporting means only, often as software agents like BDI agents. In this article, we lay our focus on a special kind of agents which are not only used as operation-supporting means in work systems, but also as part of the operating force. To co-operate as part of the operating force with the human operator on a given common work objective demands that both the human and this kind of agents are able to understand the behaviour of each other the same way as humans understand each other. This often takes a-priori knowledge of human cognition on the agent's side, explicit and implicit (knowledge of designer), in order to generate beliefs about the mental and physical state of the human operator.
As a prerequisite for that cognitive behaviour both human and agent must have got a similar understanding of how to tackle a given work objective and the explicitly existing mutual desire to comply with it. Therefore, in distinction to other known kinds of agents, we call these software agents artificial cognitive units (ACU). An ACU features its own independently derived central situation representation, thereby forming its own situational picture about all what is of interest regarding the work process at hand as it might be also available to a human operator who may work next to it in the work process. Based on that and other functions needed it acts as a kind of "one-brain artificial individual" similar to a human person. This leads to a work system architecture embodying both natural and artificial individuals as distributed cognitive entities. This is a powerful architectural feature which we will take advantage of. To make it even clearer, the characteristics of an ACU as proposed here are in strong contradiction to other approaches like a Multi-Agent System (MAS) or the Real-Time Control System (RCS) as described. An ACU is not composed of functional specialists with only a partial situational picture like we experience it as a recommended feature of individual agents in a multi-agent system. An ACU cannot be compared with an individual component of the RCS architecture. Concept-based performance according to the understanding we developed earlier, i.e. the capability of the generation of solutions to problems not previously anticipated by the system designer, cannot be easily achieved by a system reflecting a task decomposition of its application domain within its architecture.
Concept-based performance in principle needs the freedom of taking any a-priori and situational knowledge into account, no matter whether it has been previously acquired in a context which is not the same as the one actually given. What we can learn from human mastering of unknown situations and human creativity is the voluntary use of knowledge from various, probably not even directly related domains while solving a particular problem. This at least requires the architecture of a central knowledge representation within the ACU. Every design decision which breaks down the system in separate units according to task decomposition hinders the emergence of concept-based behaviours. Therefore, we advocate a design policy, which concentrates as much functionality within one ACU as possible. As an analogy from daily life, it is certainly more result-oriented to ask a well educated general practitioner for a medical diagnosis on an unclear set of symptoms first, instead of consulting various specialists, each of them having a heavy focus and preoccupation due to their special discipline.
Our website is not responsible for the information contained by this article. Articleinput.com is a free articles resource thus practically any visitor can submit an article. However if you notice any copyrighted material, please contact us and we will remove the article(s) in discussion right away.
Note: This article was sent to us by: Zane Gibbs at 01152010
1. The operating force is the most important component of the work system
All articles are property of their respective authors. Please read our Privacy Policy!
© 2009 ArticleInput.com.