This paper presents a methodology for the design of hybrid switching systems. Hybrid switching enables both circuit and packet-switched services to be provided in the same communication network. A model for the analysis of blocking and delay of a hybrid switching link is developed. The capacity assignment problem for hybrid-switched networks is also formulated and solution techniques developed. These models are incorporated into an overall hybrid switching network design methodology, and applied to study alternative system operations.

This paper presents quantitative results of cost/performance studies of circuit-switching, packet-switching, and hybrid (circuit/packet) switching technologies. The results expose the cost difference for alternative realizations and usage of a given network technology.

A store-and-forward broadcast radio network with mobile switching nodes is considered. The problems of monitoring connectivity changes in the network and updating information in switching nodes to enable routing of messages are addressed. Analytical models are developed for determining the fraction of time that switching nodes have correct information as a function of trajectory and speed for mobility, monitoring rate, traffic load, and network topology. The problems are parametrically investigated and results of monitoring rate to obtain performance requirements are obtained. These studies show that the monitoring rate is very sensitive to velocity; specifically, as the velocity doubles, so does the monitoring rate. These results are a first step in obtaining meaningful operating parameters to be used in experimental networks.

Recent studies have shown that packet radio networks (PRNET) can easily be saturated and become unstable unless efficient routing and flow control algorithms are used. To enable point-to-point packet transportation, network stations assign a code (label) to each repeater; the process of assigning such labels is referred to as "network initialization." The initialization procedure assumes that the network topology is not known a priori and is changing with time. Thus, the initialization procedure involves mapping of network topology, determining network structure (labels for repeaters), and transmitting labels to the repeaters. Although many authors have developed analytical models to study channel access schemes, no one has yet addressed directly the initialization issue. In this paper we present two models which enable one to estimate the optimal rates at which repeaters and stations must transmit connectivity information to minimize the network initialization time in a one-hop network. The first model assumes complete interference (all repeaters can communicate directly with each other), while the second (but more complex) model can provide initialization time under partial interference pattern. With these models we illustrate the effects of varying repeater and station transmission rates for small networks. Our most important results are that the network initialization time is relatively insensitive to the station transmission rate, but the repeater transmission rate must be carefully chosen to achieve rapid initialization.

A Packet Radio Network is a store-and-forward packet system employing radio links. A Mobile Packet Radio Network generalizes this system by allowing every radio unit to be independently in motion. In this paper a model for the analysis of connectivity as a function of time in such a network is presented and used to solve for parameters of interest. The model and the related analysis are guides for developing procedures for monitoring connectivity.

The problem of acknowledgment schemes in store-and-forward packet-switching communication networks is addressed. The performance of systems with end-to-end acknowledgment (ETE Ack) and hop-by-hop acknowledgment (HBH Ack) are compared in terms of packet delay and channel utilization. It is shown that a system with an HBH Ack scheme performs significantly better in terms of delay or channel utilization, when the number of hops that a packet traverses is large and/or when the probability of successful packet transmission over the channel is low.

Packet-switching broadcast radio networks are receiving considerable attention as a feasible solution for applications involving fast network deployment requirements, inaccessible physical environments, and mobile communication devices. Such networks also offer economic alternatives to traditional multiplexing schemes for local distribution. Most of the published papers relating to packet-switching broadcast radio networks address the case in which all communication devices are within an effective transmission range of the destination receiver, thus forming a single-hop network in which no packet routing is involved. In this paper, we address multihop networks. The problems encountered in packet transportation are identified and strategies to resolve these are proposed.

Packet switching over broadcast channels with random access schemes is of current interest for local distribution system and for satellite channels. This mode of operation is useful when the communicating devices are mobile and when the ratio of the peak to average data rate requirement of each device is high. Such systems have been analyzed for the case in which all communicating devices are within an effective transmission range of each other; either directly or through the satellite. In this paper, we address broadcast networks in which originating devices cannot directly reach the destination receiver. Thus, devices are introduced which receive these packets and repeat them to the destination. The capacity (maximum throughput) of such systems is determined, and design problems related to the number of repeating devices and the usefulness of directional antennas are resolved.

An algorithm for classifying a data set into an initially unknown number of categories is presented. It is composed of procedure for selecting initial points, a mode estimation procedure, and a classification rule. An integer valued function is defined on the sample space and a gradient search technique is used for estimating its modes. A procedure for mode estimation in the case of an infinite data set is also proposed. Sufficient conditions for the convergence to the neighborhood of the modes have been stated. The algorithm was used for clustering multicategory artificially generated data sets and was compared with an optimal classification scheme.

An algorithm is presented which partitions a given sample from a multimodal fuzzy set into unimodal fuzzy sets. It is proven that if certain assumptions are satisfied, then the algorithm will derive the optimal partition in the sense of maximum separation.