Showing papers on "Cognitive network published in 1989"

Rule-based network design: application to packet radio networks

Abstract: A heuristic, rule-based approach to network design is presented. The problem of designing high-performance packet radio networks by altering individual radio powers (i.e. by altering network connectivity) is considered. It is assumed that node locations, traffic requirements, and a multiaccess protocol are specified. The focus is on the dependence of capacity on connectivity for two reasons: connectivity is always a design parameter, whereas radio locations, protocol, and traffic may not always be; and the relationship between capacity and connectivity is well documented in the literature, even though the exact nature of this reliance remains unclear. Packet radio performance analysis models found in the literature are used for performance evaluation. >

A Cognitive Model of Human-Computer Interaction in Naval Air ASW Mission Management

Abstract: : Few, if any, techniques exist for cognitive modeling of human-computer interaction (HCI) in real-time multi-tasking (RTMT) problem domains, despite the fact that this class of problems includes many of the most critical and challenging person-machine systems. A new modeling formalism for representing HCI in RTMT domains is developed and applied to a realistic domain from Naval aviation, Airborne Antisubmarine Warfare (Air ASW) mission management. The formalism is called the Cognitive Network of Tasks or CCGNET model with the blackboard architecture. The COGNET technique is used to model HCI in the example domain (Air ASW), resulting in a detailed and computable model of the strategies used by the human decision-makers. The model is built from data collected from expert human operators in a realistic experimental environment, described in a previous report.

Packet radio network design

Abstract: In broadcast networks such as packet radio networks, network capacity depends on network connectivity, yet the exact M~UE of the capacity/conoectivity relationship defies simple categorization. This paper introduces a general method capable of improviag network capacity by altering connectivity values, and remains sufficiently broad to allow use on a wide range of packet radio network topologies, traffic requirements, and protocols. The approach uses artificial intelligence techniques to formulate design rules which quantitatively describe individual aspects of the capacity/connectivity dependence. Incorporation of the design rules into a rule-based computer optimization program produces, on average, 15 - 25 % capacity improvement for both random and regular network inputs. Optimization results lead to conclusions about efficient network topology and operation, and permit further capacity/connectivity generalizations. The approach used, as well as many of the rules developed, go beyond the domain of packet radio networks, and are applicable to many general network design problems.

Alternative approaches to time critical communication

Abstract: It is becoming apparent that in their present form the OSI standards cannot readily accommodate time-critical applications and that major changes will be required if real-time applications are to be successfully implemented on OSI networks. It is now becoming apparent that initiatives to define ideal architectures capable of achieving particular absolute performance will never solve the problems of time-critical communications in all applications. Therefore it is relevant to reappraise the role of the network in time critical communications and provide guidelines for integrators to assist them in achieving an appropriate implementation for a specific application.

Packet-switching network of dynamic architecture based on artificial intelligence

Abstract: A method of designing packet-switching network of dynamic architecture based on artificial intelligence is discussed in detail. According to this method, the intelligent information about the user's requirements and network states is used to control the network architecture's exchange so that its information transmission efficiency is increased greatly. Some technical problems and experimental results concerning this dynamic-architecture packet-switching network (DAPSN) are examined. It is suggested that the design principle of DAPSN can be widely applied in many computer networks, for example, the MAN, WAN, LAN, (metropolitan, wide, and local area networks), andISDN (integrated services digital network). >

Cognitive representations of DOS commands as a function of expertise

Abstract: A study was made of the cognitive networks derived from the similarity rating of IBM Personal Computer Disk Operating System (IBM PC DOS) commands by computer users with varying levels of expertise. The networks of naive, novice, and intermediate users were examined to determine which links in their networks were also present in an expert network. Groups with a greater level of DOS expertise had more links in common with the experts. A core set of commands was identified that were linked in every network. As the level of expertise changed, it was possible to show the order in which links in the experts' network appeared in those of the novice and intermediate groups. >

The coming of a savior?-network management with OSI and POSIX

Abstract: A description is given a number of tools and models which use a centralized view of network management. It is shown how this type of model is being integrated into new management systems which are being defined, such as IEEE POSIX 1003.7. The OSI (Open Systems Interconnection) model, in conjunction with the work of POSIX, provides for a complete network management model the encompasses the low-level interface, the high-level interface, and interoperability, and allows for flexible administration of systems on a network. The integration of the two provides for a seamless integration of system and network, so that administration of a network, rather than just the administration of individual machines on a network, is possible. >