ES2361301T3 - ROAD-FREE ROAD OF INTEGRATED COLLISION IN CROSSED LAYERS. - Google Patents

Abstract

A method for configuring the connection in a multi-hop wireless communication network, said method comprising the steps of: - determining, for a requested connection between a source node and a destination node, a set of connection parameters that include the path , the channel and at least one physical link parameter expanding a tree directed with the preliminary paths for the pending connection originating from the source node and performing a search procedure to find at least the cost Ki for each node i, in a set given, from the source node according to the following nested algorithm: ** Formula ** where i ≠ Source ID, N (i) is the set of current neighbors of node i which in turn is a set of all Ω nodes in the network, j is a neighboring node belonging to N (i), m is a set of at least one channel in a set of M orthogonal channels in total, Ψ is one or a multitude of physical layer parameters, ki ( j, m, Ψ), ta m well indicated kij (m, Ψ), is the cost from node j to node i, in which the cost kij (m, Ψ) includes the transmission power of the link Pj (m) for node j and channel m as a parameter of the physical layer Ψ, and the transmission power of the link Pj (m) is subject to restrictions that restrict the transmission power of the link for a predetermined interval, and the term K (j), also indicated Kj, is the accumulated cost from the source node to node j, and Source KID is the initial cost in the source node, at which the innermost level of nesting of said nested algorithm tunes to said parameter (s) of the physical layer Ψ, the next level of nesting is a choice of a set of channel (s) m for each neighbor, and the third level of nesting provides a choice between neighbors j, hence choosing the path in the routing layer; and - establish the required connection based on the determined set of connection parameters.

Description

TECHNICAL FIELD OF THE INVENTION

The present invention generally relates to Quality of Service (QoS) support in communications networks such as multi-hop wireless networks, and more specifically to the determination of connection parameters, connection configuration as well as connection admission control. in such networks.

BACKGROUND

When routing is applied in a wireless network, such a network is often indicated a multi-hop network. In a multi-hop network, nodes out of reach between them can benefit from intermediate nodes that can send their messages from the source to the destination. Traditionally, multi-hop networks have been associated with so-called ad hoc networks in which the nodes are mostly mobile and there is no central coordination infrastructure. However, the idea of multi-hop network connection can also be applied when the nodes are fixed. One such scenario points to Internet access in rural areas and uses fixed nodes attached to the upper part of the roofs of houses, lampposts or similar.

Although some research has been ongoing in the multisalt area since the early 1970s, relatively few of these research efforts have been directed towards the provision of QoS for multisalt networks. The reason is that such QoS support in multi-hop networks is considered to be of immense complexity. Unpredictable mobility, traffic patterns that change randomly seemingly, unreliable wireless channels, calculation complexity as well as other harmful effects are the cause of this scenario. However, some researchers have tried to face the challenge of QoS for multisalt networks. The most interesting and promising research in this area has focused on using some type of TDMA (Multiple Access by Time Division) spaced as the MAC structure (Media Access Control) as the basis.

The state of the art with respect to multisalt networks that provides collision-free channels, which allow QoS routes to be established between a source node and a destination node, will now be described below. These kinds of protocols are often referred to as QoS routing protocols. Unlike general routing protocols, QoS routing not only requires finding a route from a source to a destination, but the route must also meet the requirements of end-to-end QoS, often given in terms of width bandwidth and / or delay, for example to support multimedia communication in real time. The state of the art of the routing protocols of the QoS could be divided into two distinct groups, from now known as routing schemes and access to the separate channel and routing schemes and access to the integrated channel. In the first group, the task of routing and assigning the channel is separated into two different algorithms, that is, first a route is found and from then on the assignment of the channel is performed, while the second group adopts a more or less integrated approach for routing and channel assignment.

For a better understanding of the QoS routing schemes of the prior art, it may be useful to start with a brief description of the OSI (Open Systems Interconnection) model for network connections. The OSI model includes seven different protocol layers: the physical layer (1), the link layer (2), the network layer (3), the transport layer (4), the session layer (5), the presentation layer (6) and application layer (7). The physical layer, which refers to the physical aspects of network connections such as transmission media, transmission devices and data signals, is sometimes not seen as a protocol layer. For simplicity however, all layers will be known as protocol layers. Among other things, the link layer establishes and maintains the links between the communication devices and controls access to the network medium. The main responsibilities of the network layer include switching, routing and gateway services. The transport layer is responsible for distributing the frames between the services on different devices. The session layer manages dialogue control and session management. The presentation layer is responsible for the presentation of the data and the application layer refers to the provisioning of services on the network and provides an interface for applications for network access.

Routing schemes and access to the separate channel

Separate QoS routing protocols use generic QoS measurements and do not tune to a particular MAC layer, ie layer 2. To be able to ensure that the QoS requirements are met, these protocols have to be improved with a MAC protocol that provides access to the collision free channel.

INSIGNIA

INSIGNIA [1] is an end-to-end IP-based in-band signaling framework to provide QoS in ad hoc networks. In-band signaling assumes that each package carries all the information necessary to establish a reservation. The QoS mechanism is independent of both the ad hoc routing protocol used (reference is made for example [2] or [3]) and the technology of the link layer, although the final received QoS will depend strongly on these features. The operation of the fabric can be described as follows: A route from the source to the destination is found by the ad hoc routing protocol in layer 3. Since each packet carries the necessary information to reserve the necessary bandwidth, the packets of Data can begin to cross the route as soon as it has been established which leads to rapid booking. When a node on a route from the source to the destination receives a packet from a flow for which it has not reserved capacity (indicated by a bit in the header), it reserves the required capacity if possible.

Ticket Based Exploration

As was the case with INSIGNIA, Ticket-Based Exploration (TBP) [4] is a pure Layer 3 protocol in which all the signaling on this layer is performed and that needs Layer 2 (MAC) support to decide If a reservation can be accepted or should be rejected. TBP however is a true ad hoc routing protocol. The main purpose in reference [4] is to locate the search for feasible paths between the source and the destination for only a part of the network instead of flooding the entire network as usual in ad hoc routing protocols. More specifically, they want to search only a small number of roads from the source to the destination, instead of doing a thorough, expensive search. This is achieved by issuing tickets. A ticket is the permission to search for a road and from there, the maximum number of roads searched is limited by the number of tickets. When an intermediate node on the way from the source to the destination receives a ticket you have to decide to which node (s) the ticket should be sent. To do this, the node uses the status information to guide the limited packets along the best routes. A distance vector protocol is used to collect this status information consisting of end-to-end delay, bandwidth and cost.

Examples of conflict-free programming algorithms

In [6], Nelson and Kleinrock introduced the concept of Spatial TDMA (STDMA), where time intervals (TS) are spatially reused. This work can be considered as the parent of all other programming algorithms for the purpose of providing conflict-free programming. The idea is to determine the sets of non-interfering (or non-colliding) links. This involves a stationary network, and the assemblies need to be recalculated if the network changes sufficiently. These sets are preferably selected such that each node in the network is allowed to transmit at least once. Each time slot in a TDMA frame is then assigned a set of links (transmission sets) that can transmit without interfering with each other. The same programming is repeated every STDMA frame later.

The schemes presented in [8] and [9] could be seen as a direct descendant of STDMA. In previous work on STDMA, network chart connectivity is used to decide if interference occurs. Such an approach does not capture total interference in the network. Instead, in the scheme (s) of [8] and [9], the generation of STDMA programs is based on the value of the SIR for each link.

STDMA and its descendants are more or less designed for centralized or offline calculation of transmission sets. In ad hoc networks, this is not a sensible approach.

A particular layer 2 scheme, which ensures distributed and collision-free operation, called Collision Free Topology Dependent on Channel Access Programming or CTMA is presented shortly in [7].

“Integrated” routing and channel access schemes

The “integrated” QoS routing algorithms are based on the assumption that the ad hoc network can be modeled as a graph, and all of them are protocols on demand, that is, the route search is done only after it has been a route from a higher layer protocol has been requested. Additionally, the integrated QoS routing algorithms generally involve TDMA, fixed transmission power and omnidirectional antennas. In a graphic, two nodes are called neighbors if they are able to communicate with each other and this is represented by connecting the two nodes with a link in the graphic model. Two nodes are connected if the distance between them does not exceed a predetermined value; that is, packets are received without error in the absence of external interference from other nodes. It is also assumed that only the neighbors are interfering with each other. In a multi-skip packet radio network modeled with graphics, transmissions are often modeled to interfere in two ways, hereafter known as primary interference and secondary interference. Primary interference occurs when the node is supposed to do more than one thing in a single time interval, for example transmit and receive in the same time interval. Secondary interference occurs when a receiver R tuned to a particular transmitter T is within the range of another transmitter whose transmission, although not intended for R, interferes with the transmission of T a

A. When a graphic model is used, it is sufficient to prevent transmitting to all the neighbors of R in the same time interval as T to avoid secondary interference. When several integrated QoS routing protocols are described below, they have been classified according to what level of interference they are considering

- the interference-free channel, only considered the primary interference and considered both the primary and secondary interference, since this greatly affects the way the protocol is designed.

Most existing ad hoc protocols only refer to the existence of a shorter path route between two nodes in the ad hoc network without guaranteeing their quality. As previously described, the purpose for routing protocols of ad hoc QoS is to configure a channel / path from a source node to a destination that meets some requirements regarding bandwidth and / or delay to be able to support Multimedia communication in real time. To do this, conventional integrated QoS routing protocols typically consider bandwidth in a link when looking for a route from source to destination. The bandwidth requirement is then made by reserving time intervals on the links along the way. The main advantage of this approach, when compared to ordinary ad hoc routing protocols, is that the requirements of the QoS can be met. Compared to the completely separate QoS routing protocols, this means that QoS provisioning can be achieved with better network utilization.

To calculate the available bandwidth in this environment, it is incorrect to simply calculate the minimum bandwidth of the links along the way as is done in wired networks. The cause of this is that the available bandwidth is shared between adjoining nodes. A simple example of this is the following where only primary interference is considered: Suppose node A wants to communicate with node C through node B. The free intervals available for A to communicate with B are 1, 2, 3 and 4 and the same is true for the link from B to C. If this had been a wired network (or an interference-free channel) the available capacity would have been 4 intervals, while in this case the capacity is 2 intervals. The reason for this is that the intersection of common free intervals on links A to B and B to C is not an empty set and a node is not able to both transmit and receive in the same interval. In addition, we assume that the free intervals available to communicate from A to B and from B to C are 1, 2, 3 and 4 as well as 3 and 4 respectively. If intervals 1 and 2 are reserved for communication from A to B and intervals 3 and 4 for communication from B to C, the available bandwidth from A to C is 2. On the other hand if intervals 3 and 4 Reserved for communication from A to B will mean that no communication from B to C can take place and the available bandwidth from A to C would have been 0 in this case. The protocols that are able to identify this problem and solve it will be said to be able to perform optimal programming from now on. There are two problems involved in this process of calculating the bandwidth of the road. The first problem is how station B (here it is assumed that B is responsible for reserving capacity on the link from A to B) knows the set of common free intervals of two adjacent links, and the second problem is how to share this information with its neighbors. To solve these problems, the station has to exchange some messages between them.

Interference Free Channel

Reference [16] describes a multipath QoS routing protocol that is based on the bill-based approach presented in [4]. The term "multipath" refers to the case where the capacity reserved from the source to the destination can be divided into several subcategories, each serving part of the original required capacity. However, this work is quite an ideal model in which the bandwidth of a link can be determined independently of its adjoining links. To support this assumption, it is assumed that each host has multiple transceivers that can work independently of each other and that each link is assigned a code that is different from those codes used by its two-hop neighbors to avoid collision. A two-hop neighbor is a neighbor of a neighbor - in the previous example A and C are two-jump neighbors.

Considered only Primary Interference

An assumption less strict than the interference-free channel is made in [12], [13] and [15], where a CDMA channel model is assumed over TDMA, which implies that the use of a time slot in a link is dependent only on the state of its adjoining links (that is, they only consider primary interference). The focus on these three references is the calculation of the bandwidth available on the way from the source to the destination but the way in which the required information is collected differs.

The general operation of [12] and [13] will now be described briefly. Upon receiving a RREQ (Resource Request) the bandwidth from the source to this node is calculated. The bandwidth can be calculated in an optimal way since the information is exchanged with its neighbors about the available free intervals prior to the calculation of the available bandwidth and that the RREQ message includes the intervals used for the previous link in the path from The node to the source. The RREQ falls if the result does not satisfy the QoS requirement. As expected, the destination will receive more than one RREQ each indicating a unique feasible path from the source to the destination. The destination will choose one of the roads and issue a RREP (Route Response) message. As the RREP message crosses back to the source, each node along the path reserves those free intervals that were calculated in advance.

In [5], a protocol for the routing of QoS in an IEEE 802.11 network using the bandwidth calculation algorithms described above is presented.

In [15], instead of calculating the available jump-to-jump bandwidth, each RREQ packet records all the link status information from source to destination. In this way the destination is able to calculate the best path from the source to the destination and issues an RREP message along the chosen path. An option for multipath routing is also presented in the reference and is easily achieved since the destination has all the information of all available paths from the source to the destination. The algorithm proposed in [15] points to a network of flows; that is, it supports multiple different flows. It is the task of the destination node to determine the network of flows from the source that meets the bandwidth requirement. Although such a solution has the potential to provide a closure to the optimal route, with interference being omitted, it also puts immense calculation load on the destination node.

Considered both Primary and Secondary Interference

Group based

In the group-based networks described in [10], [11], [14] and [18], a node could be a group head, a gateway or just a usual node. Once a node is chosen as a group head, all its neighbors belong to the same group. A node that belongs to two or more groups plays the role of a gateway. CDMA is used to divide the group by assigning different code sequences to different groups, and TDMA is forced into a group. In doing so they demand that secondary interference only has to be considered within the group, since it is assumed that the interference between the groups is insignificant. To spread the information of the available intervals within the group each node periodically transmits a "free interval" message containing its interval reserve status. Since the head of the group can hear all the other nodes in the group, it has full knowledge of the reserve status in the group. Since the head of the group, like all the nodes in the group, is obliged to transmit the “free interval” message, all the nodes will definitely know the reservation of intervals within the entire group. This makes it simple to calculate the available bandwidth. Signaling and calculation of available bandwidth are then carried out independently on each node in a jump-by-hop manner.

The scheme proposed in [18] is not really an integrated approach. Instead, a hierarchical scheme is proposed - the first capacity allocation is done both at the node level (link) and at the flow level (in both of these steps a fixed routing is assumed), then a distributed version of the Bellman-Ford algorithm for final route selection. It is assumed that all assigned intervals of the node are assigned to the appropriate part of the time frame by some mechanism, but the MAC layer is not considered. In addition, it is required that a central entity collect all the information, perform the calculations and then distribute the final result, that is, it is a centralized approach.

In the references [17] and [19] mentioned below, an ordinary TDMA / spaced structure is assumed and hence the protocols will have to consider both primary and secondary interference with respect to all nodes in the network.

Liao

The scheme presented in [17] requires that a first node of everything has to know the partial graph of its own location (this means that a node has to know how its neighbors and neighbors' neighbors interconnect (usually known as neighbors of 2 jumps.) In addition, a node also has to have full knowledge of the intervals at which the neighbors are receiving and transmitting (pointing out, it is not enough to know if the node is busy or not.) This also holds for its neighbors of 2 jumps To create these data structures, a central computer needs to periodically disseminate this information to its neighbors and they have to redistribute it to their neighbors.With this information it is possible to conduct routing and assigning intervals.The rule is (as in most other documents) that an interval can be assigned only if the two nodes commonly indicate an interval as free and that the sending node n or interferes with any of its neighbors. Note that this scheme is not able to calculate the optimal bandwidth of the path.

Zhu

The scheme presented in [19] resembles some of the approaches that consider only primary interference, but is not able to optimally calculate the available bandwidth. Nothing is said about how a node knows at what intervals they are allowed to send with respect to secondary interference with other nodes not on the way from the source to the destination. It is merely stated that it is the job of the protocol for assigning underlying intervals in the MAC layer to determine how the nodes negotiate with each other to ensure that the intervals are assigned to the corresponding transmitters and respected by their neighbors.

Additional state of the art solutions

In [20], a graphic model is constructed assuming that two nodes are connected if the distance between them does not exceed some predetermined value, that is, packets are received without error in the absence of external interference from other nodes. A relatively realistic model of secondary interference is used. Two or more stations may transmit in the same stipulated time interval that the Signal to Interference Ratio (SIR) at all reception nodes is above a certain threshold. The routing decision is based on minimal jump connections between source-destination couples. Given a network topology (represented by the graphic model) the number of possible jumps and paths can be found by spreading a packet through the network and counting the number of nodes visited. When multiple paths with equal number of jumps between the source and the destination are the assignments of previous intervals and the relative traffic load are used as decision criteria, to consummate the load balancing in the network. By doing so congestion is less likely and the data flow can be increased. In summary, the algorithm that follows five steps can be described as follows. In the first step, the graphic model is used to deduce the network topology. Next, a routing decision is used to produce balanced traffic between links taking into account the available capacity. Since capacity is required for each link for this routing decision, the capacity of the link is considered the same in this step. In the third step, any conflict-free programming algorithm, such as [8], can be used to generate the program. After this, the routing decision (fourth step) is made again, but this time based on the actual capacity given by the program. In the final step, the routing decision (step two or four) that produces the maximum data flow for the entire network is chosen. The reference [20] is really based on sequential routing and thus separate and programming / reservation. It should also be noted that this scheme requires a centralized path and the determination of resource allocation.

References [21-22] do not relate to the issue of routing in the network layer, but rather concern adaptive wireless communication, with the parameters in the physical layer and the MAC layer that are adaptively modified by a controller Base station.

The article Power Control Routing for Multisalt Wireless Ad-hoc Network by K. Tsudaka and others describes an approach to control the transmission power to improve the performance of routing protocols.

Problems associated with prior art solutions

Routing and separate channel access schemes are generally far from optimal. The reason is simply that the problem of assigning routes and channel resources has been subdivided into two simpler problems. In addition, separate schemes often involve centralized and offline path determination and resource allocations. This means that they are relatively poor in managing mobility, since the information must be collected, processed and disseminated in the consecutive results to the nodes involved.

Although several good ideas have been presented for some form of routing and access to the “integrated” channel, important radio aspects are completely neglected. Therefore, the usefulness of the algorithms can be questioned. For example, an excessively simplistic assumption that nodes use orthogonal codes so that no transmission node interferes with any other receiving node is used in several of the documents. This is not only incorrect, since in practice the codes are not perfectly orthogonal due to, for example, the spread of the delay and hence will cause harmful interference, but it is also an inefficient use of valuable resources. The orthogonality of the signals is a result of an expansion of the bandwidth (BW), and the bandwidth could probably be used better by sending the data at higher speeds. The procedure for allocating resources and paths proposed in conventional routing and access schemes to the “integrated” channel is also very simplified and can sometimes announce routes that are not feasible in practice.

SUMMARY OF THE INVENTION

The present invention overcomes these and other disadvantages of prior art adaptations.

It is a general object of the present invention to improve the use of available resources in a communication network.

It is also an object of the invention to provide a robust and efficient mechanism for supporting QoS in communications networks such as multi-hop wireless networks. In this regard, it is desirable to exploit the full potential of the network, while ensuring quality of service.

Another object of the invention is to provide considerably collision free or non-interfering communication for each individual connection, at least in a given subset of the network.

Still another object of the invention is to provide an improved method and corresponding control system for the configuration of the connection in a communication network such as a multi-hop wireless network.

Still another object of the invention is to provide an improved method and corresponding control system for the determination of a connection in a communication network such as a multi-hop wireless network.

It is also another object of the invention to provide an improved method and corresponding control system for the admission control of the connection in a communication network such as a multi-hop wireless network.

Another object of the invention is to provide a communication network having a plurality of network nodes, at least one of which includes the means for the improved determination of a connection.

It is also an object of the invention to find ways of controlling the complexity of calculation involved in determining the appropriate connection parameters.

The invention basically proposes a cross-layer integration of the functions into several network protocol layers in a single unified mechanism by means of the integrated optimization of a single objective function with respect to the connection parameters in at least three layers of protocol.

Preferably, the protocol layers involved include the network layer, the link layer and the physical layer. However, it should be understood that other protocol layers can be used in integrated optimization. It is also possible to use more than three protocol layers in the optimization, for example considering the three lowest levels in combination with an adaptive application in the application layer. Indeed, the unified approach of the invention partially or completely eliminates the need for a layered representation. Instead of having several separate optimization algorithms running more or less independently in the different protocol layers, a single unified optimization is performed.

In a preferred embodiment of the invention, the functions of the physical layer, access to the channel, routing and optionally also admission control are integrated in a unified, unique mechanism using the connection parameters that include the path, channel and one or more parameters layer / physical link in the integrated optimization. In this case, each connection is consequently defined by at least one triplet comprising a selected path, a selected channel and one or more physical layer / link parameters.

To provide collision-free or non-interfering communication, the optimization is subject to one or more restrictions designed to ensure considerably collision-free communication with respect to existing connections as well as the required connection.

By incorporating the physical layer connection parameters into the integrated optimization and performing the optimization under one or more interference-related restrictions, the issue of interference can also be carefully considered in a unified approach to the provisioning of QoS in such networks. like multi-hop wireless networks. This means that it is possible to determine the connection parameters that ensure considerably non-interfering links, including also the links of the required connection.

In practice, the objective function may include terms such as link transmission power, delay, local charge, battery power and link margin. The physical layer parameters typically define the operation of the link and include one or more physical link parameters such as the transmission power, the modulation parameters, the bandwidth, the data transmission rate, the correction parameters of mistakes, and so on. Other physical link parameters include the multiple input-multiple output (MIMO), adaptive antenna (AA) and other multi-antenna configuration parameters, on the transmission side, the reception side, or both.

Advantageously, the integrated optimization is carried out by means of a heuristic algorithm. The connection parameters can be determined for example in a local search procedure. In this regard, it has also proved useful to work with a nested algorithm, in which each level of nesting represents a protocol layer of the network.

In a special embodiment of the invention, the horizon on which the algorithm acts is made selectable to provide optimum performance for an acceptable level of calculation complexity given.

In addition to a centralized implementation, in which a single predefined unit determines the connections on request, it is also possible to distribute the optimization algorithm to a plurality of network nodes in the network, using RREQ (Resource Request) and RREP ( Route Response) for the transfer of the required information. In the distributed scenario, for a given connection request, the optimization algorithm can be executed on the relevant network nodes in a node-to-node manner, or run entirely on the destination node using the information collated in a RREQ that has been sent via network.

The networks considered mainly are wireless networks (radio, optical, etc.) multisalt, but the invention can also be applied in other networks such as multiple access networks formed as hybrids of wireless and wired technologies.

It should be understood that the term "channel" includes time intervals, frequency bands, orthogonal codes,

or any other orthogonal channel or combinations thereof, even non-spaced channels.

In summary, the present invention concerns the provisioning of QoS in communications networks such as multisalt wireless networks, but also points to the efficient use of the wireless medium while ensuring the use of low complexity path discovery algorithms. The proposed algorithms are not limited to the stationary case of a fixed node scenario, but can handle under certain circumstances low

or moderate mobility.

The invention offers the following advantages:

-High network utilization;

-Efficient provisioning and support of QoS, including delay and guaranteed data flow;

-Communication considerably free of collisions;

-Low complexity of calculation given the increases in performance and the combinatorial complexity of the optimal solution;

-Flexible control of calculation complexity;

-

Reduced power consumption, when the transmission power is used in the objective function;

-

Reduced end-to-end delay, when the delay is used in the objective function;

-Both implementations distributed as centralized are feasible; Y

-Allows the selection of the shortest path near or, at very low load, real.

Other advantages offered by the present invention will be appreciated in reading the subsequent description of the embodiments of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention, together with other objects and advantages thereof, will be better understood with reference to the following description taken together with the accompanying drawings, in which:

Fig. 1 is a schematic diagram illustrating the routing and access schemes to the channel according to the prior art; Fig. 2 is a schematic diagram illustrating the unified approach according to the invention; Fig. 3 is a schematic diagram of an exemplary multi-hop wireless network; Fig. 4 is a flow chart of the configuration, rejection and disassembly of the connection; Fig. 5 illustrates the notation used for a preliminary connection path configuration in a network exemplary multisalt wireless; Fig. 6 illustrates a preliminary connection path configuration in a multi-hop wireless network specimen for a channel and pair of specific nodes; Fig. 7-12 are schematic diagrams illustrating an example of the operation of a mechanism unified inventive for the provisioning of QoS in a given network; Y Fig. 13 is a schematic diagram illustrating a graph of the resulting CDF CIR for all active receptors corresponding to the example of Fig. 12; Y Fig. 14 is a schematic diagram of a network node in which a CFPR algorithm of according to the invention; Fig. 15 is a schematic diagram illustrating an example of a non-spaced channel reservation.

DETAILED DESCRIPTION OF THE EMBODIMENTS OF THE INVENTION

For a better understanding of the invention, the prior art will be summarized first with respect to routing and channel access schemes with reference to Fig. 1. In conventional schemes for routing and access to the channel, each protocol layer is associated usually with its own independent algorithm A1, A2, A3. Sometimes extraction data can be sent from a lower layer to a higher layer to provide some form of sequential or "smooth" layer crossing integration. Extraction data from the lower layer is simply transferred to the higher layer for use by the higher layer algorithm, without feedback to the lower layer for adaptation. For example, the extraction data concerning the link bandwidth can be transferred from the link layer, L2, to the routing algorithm A3 in the network layer L3. The bandwidth information can then be used by the routing algorithm, which for example can change the path assignment if the bandwidth is too low. This approach of using several independent algorithms, each concerning its own objective function, represents a relatively simple form of layer crossing integration that only provides suboptimal results, if feasible paths can be produced at all. A careful analysis of conventional schemes for routing and access to the integrated channel also reveals that they all work only in two layers of protocol and that they often omit the interference issue completely and therefore provide suboptimal routes and do not exploit the full potential of networks.

References [23-25] are all US Patent Application Publications published after the filing date of the U.S. Provisional Patent Application. No. 60 / 358,370 on which the present patent application is based.

Reference [23] describes the routing for ad hoc interworking based on the link quality measures transferred from the link layer.

Reference [24] describes a cellular network with the allocation of adaptive channel resources based on a criterion of minimum distortion or minimum power, taking into account the wireless transmission characteristics that vary over time.

Reference [25] describes the use of a multimedia extraction unit for the integration of the management of the link layer with the management of the network layer. Several transmission parameters are modified in response to changing environmental factors, and this modification changes the bandwidth of the available link, which in turn is used in network layer traffic management.

References [23-25] represent merely different variants of "soft" layer crossover integration, which involves at most two protocol layers at the same time with the extraction data that is transferred from a lower layer to a more layer high.

As schematically illustrated in Fig. 2, the invention proposes a unified algorithm for the integrated optimization of a single objective function with respect to the connection parameters in at least three protocol layers, preferably including layers L1, L2, L3, thus providing a truly integrated and unified approach to routing, channel assignment and adaptation of physical link parameters. Instead of having separate optimization algorithms that run more or less independently, a single unified optimization according to the invention is performed. As previously mentioned, the unified approach of the invention can actually eliminate the need for a representation of layers, although there is nothing that prevents other optional functions, represented by the boxes of dashed lines in Fig. 2, still residing in the network layer L3, link layer L2, and physical layer L1. These functions may or may not be cooperating with the unified algorithm according to the invention.

The invention will now be described with reference to a particular communication network, namely a multi-hop wireless network (radio, optical, etc.). However, it should be understood that the invention is not limited thereto, but can also be applied in other networks such as multiple access networks formed as hybrids of wireless and wired technologies.

Basic principles and description of the network

As mentioned above, the invention generally represents a true cross-layer integration of functions in several protocol layers in the network, thus providing a unified approach to the provisioning of QoS in a multi-hop network. In the unified approach according to the invention, the connections are preferably determined by the integrated optimization of a given objective function with respect to the connection parameters in at least three protocol layers within the network.

For a better understanding of the invention, it will be useful to begin with a brief description of an exemplary multi-hop wireless network. Fig. 3 is a schematic diagram of an exemplary multi-hop wireless network that illustrates some important concepts. We are supposed to have a network comprising multiple nodes. The network operates on a wireless medium in which transmissions can potentially interfere with each other. Traffic sent between two nodes in the network is called flow. The sender of the data in such a stream is called the source node and the receiver is called the destination node. At every moment, the network transports zero, one or a multitude of traffic flows. Each flow is transported in a connection (also known as a circuit from the classic network connection). For simplicity, only a single connection is shown in Fig. 3. In practice, of course, several simultaneous connections can exist.

A connection from a source node to a destination node is defined by a series of connection parameters, which include the path, as well as the channel parameters and the physical layer parameters along the path. There may be several different paths from the source to the destination. Each path is mounted by a set of links, and each link between two adjacent nodes i and j can use several different communication channels and physical link parameter settings. A path can be characterized by the identities of the nodes.

The physical layer parameters, also known as physical link parameters, can be associated with the transmission side and / or the reception side of each node along the path. The parameters of the physical link for the transmission can be for example the transmission power, the modulation and so on. The link parameters for reception may include tuning the antenna groups. Communication on separate channels is supposed to be entirely orthogonal and hence cannot interfere with each other. The change from one channel to another in a relay node is called channel switching. A connection typically has an upper limit of the data transmission rate, and a stream can use a fraction of the available data rate or full bandwidth. The nodes within reach between them are generally said to be neighbors. Obviously, several definitions of the term "within reach" are possible. Preferably, the condition of being within the range of a node is that the Average Signal to Noise Ratio (SNR) in reception exceeds a predetermined level when the maximum transmission power allowed in the sending station is used and there are no interfering stations.

5

fifteen

25

35

Four. Five

It is desirable to determine the connection parameters that are optimal in some sense. To be able to talk about optimality in a well defined way, an objective function f is introduced. In general, the objective function f is carefully selected and made dependent on the connection parameters such as the path, the channel and the physical layer parameters. Even though the parameters of the physical layer / link are normally part of the objective function, other factors of the non-physical layer / link, such as the local charge or the remaining battery, could be incorporated into the objective function. The objective function is then optimized with respect to the relevant connection parameters, thereby jointly determining the optimal connection parameters for a connection. The term parameter is used both to represent a variable parameter as such, as a variable parameter value, as quickly understood by experts.

When the optimization is formalized, the following notations can be used:

Ω indicates all nodes in the network (or the considered part of the network).

M indicates the set of orthogonal channels in total.

Ψ indicates one or a multitude of parameters of the physical layer, and in this way it can be multidimensional with respect to the parameters of the physical layer, each variable parameter as such having a definition space in which it can assume continuous or discrete values.

In optimization, the objective function f, which uses the input parameters of the previous sets Ω, M, and Ψ, obtains the actual connection parameters that define the connection, including the path, the channel and the physical layer parameters :

R defines the real nodes between the source and the destination:

image 1

Su defines the set of channels that the node uses for transmission:

image 1

where M (u) is the set of optimal channels for node u, and node u belongs to path R, except the destination node.

Tu, v defines the set of parameter values for node u in channel v, and may include the transmission and / or reception parameters, except for the source node (transmission parameters only) and the destination node (only reception parameters):

image 1

where Ψ (u, v) is the optimal set of parameters in node u for channel v, and v belongs to M (u), and node u belongs to path R.

The fact that a single objective function is optimized with respect to the path, channel and parameters of the physical layer really causes a true cross-layer integration of routing in the network layer, the assignment of the channel in the link layer as well as the functions of the physical layer in a single unified mechanism.

When subjected to appropriate design constraints, the cross-layer optimization proposed by the invention causes a network that has one or a multitude of connections assigned in such a way that considerably collision-free communication is guaranteed for each individual flow. In reality, completely collision-free communication is not possible. However, collision immunity can be defined as practically maintaining any relevant performance measure such as PER (Package Error Ratio), CIR (Carrier to Interference Ratio, assuming noise is included) or SNR (Relationship a Signal to Noise) below certain target values or within predetermined target intervals. For example, collision-free communication can be defined as being fulfilled as long as packet losses caused by interference and other harmful radio effects remain arbitrarily small.

The optimal solution to such layer crossing integration is very complex computationally. For example, only the selection of the path and channels in a multiple access scheme such as TDMA or FDMA is well known to be a problem called full NP (the execution time is not polynomial, that is, it is often indicated as exponential, in the size of the entrance). The incorporation of additional functions such as the optimization of physical link parameters complicates the matter even worse.

The overall strategy is to manage the provisioning of the QoS and the admission control in a per-flow manner, similar to that described for conventional integrated routing and channel access schemes. However, several new additional aspects are also taken into account here.

The general strategy will now be described with reference to the flowchart of Fig. 4. When the first flow is to be established in a network on a configuration request (S1), a connection that is optimal in some sense should be selected. Preferably, this involves selecting a path, channel and a set of physical parameters such that the connection is free of collisions (ie not affected by its own transmission) and optimizes a predetermined metric (S2). The connection admission control can be exercised using information on whether or not a feasible path is found (S3). If a feasible path is found, the connection (S4) is established and the data is sent (S5). For each additional flow that is established, the procedure is repeated, but it is also ensured, with high probability, that the new connection does not cause collisions for or experience collisions from the existing connections. In the event that a required connection cannot be configured due to the restrictions to which the optimization is subject, the flow will not be supported in the network (S6). User action in such a case is not the main focus of the invention, but it could typically involve a restarted connection configuration with lower QoS requirements (ie reduced data transmission rate if supported) or a postponement of the configuration at a later time when the network load may be lower. An additional alternative is that the destination may have been able to deduce the information on the maximum QoS available during the configuration phase. This information can then be sent to the source to guide it in a new configuration. If allowed by the source, the destination could also reserve a path that does not fully meet the optimal QoS requirements.

After a request for data release (S7), the sending of data (S8) is terminated. When the flows are finished, the corresponding network resources and the wireless medium (S9) are released, thereby leaving space in the middle and increasing the likelihood that new connections can be established whenever needed.

In the process of determining a connection, the physical layer parameters are preferably selected in this way so that a sufficient margin for reliable operation can be maintained, while ensuring that the existing connections as well as the preliminary part of the new connection do not interfere. Preferably, the physical layer parameters are physical layer parameters, which select the transmission and / or reception parameters for each link. Hence, the adaptation of the transmission and / or reception parameters can therefore affect the path taken. Conversely, the path taken affects the adaptation of the transmission and reception parameters.

A connection on demand is configured in this way whenever required, with all connection parameters such as {the path, the channel, and the physical link parameters} that are selected wisely to ensure non-interfering and non-colliding links, which include the link chain itself of the preliminary connection. In other words, in this embodiment of the invention we are dealing with the configuration of the event-driven connection with selection of the path, channel and link parameters.

Various data transmission speeds in the network can also be supported as well as mechanisms that allow the adjustment of the data transmission speeds. An important advantage is the efficient use of resources, which can be interpreted as a reduced blocking probability for any given load.

Depending on the selected objective function, optimization can be a minimization or maximization. In a preferred embodiment of the invention, the general objective of optimization is to minimize an objective cost function.

Advantageously, the optimization is carried out by means of a heuristic algorithm, for example by determining the connection parameters in a local search procedure. More specifically, the problem of the objective function can be formulated as a coupling of the specific objective functions of the node. It has also proved useful to work with a nested algorithm.

The invention will now be described with reference to a particular example of a nested optimization algorithm aimed at minimizing cost.

Example algorithm for the determination of the path, channel and link parameters

The idea is to cover a tree directed with the preliminary paths for the pending connection originating from the source node. Once the algorithm has stabilized, the route connecting the source and destination node is selected. The algorithm therefore includes a search procedure to find the lowest Ki cost for each node i, in a given set, from the source node designated according to the following algorithm, generally indicated CFPR algorithm (Collision Free Path Routing) :

image 1

KID Source = constant

where i ≠ Source ID, N (i) is a set of current neighbors of node i which in turn is a set of all Ω nodes in the network, j is a neighboring node that belongs to N (i), m is a set of one or more channels in a set of M orthogonal channels in total, Ψ is one or a multitude of parameters of the physical layer, ki (j, m, Ψ) is the cost of node j to node i, and the term K (j) is the cumulative cost of the source node to node j. The cost ki (j, m, Ψ) implies a value for each channel and some (s) parameter (s) of the physical links. Source KID is the initial cost at the source node and assumes a constant value that is typically set to zero. The set N (i) is a selectable set of the current neighbors of node i, and they do not necessarily have to include all the neighbors

The set m of channels may include one or more channels, depending on the bandwidth required. For narrowband connections, a single channel may be sufficient. For broadband connections, several channels may be required.

Since it is commonly accepted to write indexes of the nodes, such as i and j, as subscripts we will simply indicate ki (j, m, Ψ) as kji (m, Ψ) and K (j) as Kj. With this notation, the CFPR algorithm, aimed at finding the lowest Ki cost from the source node, can be summarized as follows:

image 1

KID Source = constant

Each level of nesting in the equation represents approximately one protocol layer. The innermost argument tunes the parameters of the physical layer, such as the transmission power. Hence, kji is typically a cost that depends on the value assumed by the parameters of the physical layer (Ψ), but other factors of the non-physical layer could also be incorporated. Such factors could be for example the local charge, or even the remaining battery capacity. The next level of selection is a choice of the best set of channel (s) (m) for each individual neighbor. This represents access to the MAC channel or level. The third level provides a choice among the neighbors, hence choosing a path in the routing layer.

In general, the optimization can be completely centralized to a single predefined unit, stipulating that the required information is either known or can be collected with the execution unit. However, it is also possible to distribute the algorithm to a plurality of network nodes, preferably all network nodes within the multi-hop network. In the distributed scenario, the detailed implementation depends on whether complete information or only local information is available. In the latter case, the algorithm can be executed in the relevant network nodes in a node-to-node manner, preferably using RREQ and RREP signaling for the exchange of the required information, as will be described in more detail below. On the other hand, if the required information is collected on the way to the destination node, the CFPR algorithm together with a corresponding admission control procedure can be executed entirely on the destination node based on the information collected.

The CFPR algorithm defined above is particularly suitable for distributed implementation, in which the optimization algorithm is distributed to a plurality of network nodes, and is executed consecutively in the network nodes involved in a node-to-node manner. This generally assumes that a local search procedure is performed on each node i to evaluate the cost kji (m, Ψ) from node j to node i for all nodes j in a selected set N (i) of neighbors, and determine the lowest cost Ki for each node i from the source node based on this evaluation along with the information about Kj received from each adjoining node j. The search procedure continues node by node until the entire tree of the relevant nodes has been extended.

Note that the CFPR algorithm shows some similarity to the Bellman Ford shortest path algorithm. Do not

5

10

fifteen

twenty

25

30

35

40

Four. Five

fifty

However, there are several differences. The CFPR algorithm is generalized to multiple dimensions (channels), provides an integrated optimization of the physical layer, channel and path parameters, guarantees collision immunity and provides a shorter path closure. In low load situations, the road taken will normally be the shortest real road. Fig. 5 attempts to visualize some terminology and notation used in connection with the CFPR algorithm.

Now, when a new path is set up, it is desirable to consider the links of the existing connections and avoid interfering with them. Similarly, it is also desirable to be sure that the links of the existing connections do not interfere with the links in the new connection. Therefore it is useful to divide the nodes into four sets, focusing only on two nodes at this time, that is node i and j. The first two sets are simply the node that is intended to be received, that is node i and the node that is intended to be transmitted, that is node j. A third set is a set of neighbors of node j, N (j) but excluding node i. The nodes within this set are indicated u. The fourth set includes the neighbors of node i, N (i) but excluding node j. The nodes within this set are indicated v. Typically, adjoining sets have approximately the same set of nodes.

As indicated above, any suitable cost function can be used in the optimization, and the type of physical connection parameter (s) selected may vary depending on the detailed objective of the optimization. However, for a better understanding, an illustrative example of an optimization involving physical link parameters will now be described.

Optimization example involving physical link parameters such as transmission power

According to a preferred embodiment of the invention, collision immunity is ensured by selecting the transmission power for node j such that sufficient γ reception margins can be granted for the nodes involved. For reception at node i, node j should use a transmit power Pj (m) such that the resulting reception power Ci (m) exceeds the interference level, generated by nodes v (alternatively, the other nodes at Ω can also be used in the calculations) within the set N (i) \ {j} and seen at node i, with a mitigation factor γ M. Similarly, node j should use a transmit power Pj ( m) such that it is at least one reception factor γ R less than the reception power Cu (m) at any of the nodes or within the set N (j) \ {i}. Here, the gain matrix of the G (m) channel is assumed to be known and with that it is possible to relate the reception and transmission powers between them. In addition, the noise level W uses a γ W factor to ensure that one is generally limited by interference rather than limited by noise. An additional and important condition that must be met is that node j does not interfere with the links along its own preliminary route to the source node. R indicates the set of nodes along the preliminary path connected to the node jyr indexes the nodes with R. Finally, the noise level of the receiver is assumed to be W. The maximum permissible and minimum transmission power required from the node j can now be defined as:

image2

respectively for a node within the set R. Pv (m) and Cu (m) on the other hand they indicate the transmission and reception power respectively for the nodes with established traffic. Later, when the algorithm has converged, the preliminary path that connects the source and the destination will be selected and established as an active path until its validity expires. All transmission powers as well as reception power levels will be updated to reflect the newly established connection. image 1

Note here that the formulation of is such that it does not guarantee that the resulting CIR in an existing receiving link node will not deteriorate below a certain level of the CIR. Instead it represents a simplification in which the CIR is unlikely to degrade significantly below image 1 γ R, stipulated on

mitigation margin γ M> reception margin γ R. The situation when

 It is selected such that the CIR ensures that it does not degrade below a desired CIR level will be described later.

Since some channels are not used to transmit or receive, the algorithm formulation requires that the transmission power be set to 0 and the reception power to ∞. In practice, one does not need to consider such channels when cost calculations are made and consequently one can skip them.

5

10

fifteen

twenty

25

30

35

40

Four. Five

fifty

55

A reasonable cost metric for kji is the transmission power of the link. Such a metric chooses to minimize the cumulative transmission power used on an entire path. This is good for reducing battery consumption, but it also reduces the level of interference in the system leaving space to incorporate new connections. In this way the system can operate at a higher network load. The metric is selected as:

image 1

where image 1 C is a constant selected in the interval between 0 and 1. This assumes that image 1 kji image 1 is restricted by

image 1

Y . For C = 0, kji is equal to. For C = 1, kji is equal to. The reason for adjusting the cost to ∞ is that the cost kji (m) would only assume a useful value when it is feasible.

For the correct and rapid convergence, provided that Ki = ∞, node i will discard the preliminary path leading from the source and set the relevant transmission power to 0 and the relevant reception power to ∞. Any node that has node i in its preliminary path will repeat the procedure.

Fig. 6 shows a preliminary path from node 3 (the source node) at ch 2 → node 6, channel switching ch 2 → ch 3, node 6 at ch 3 → node j. The situation shown reflects the test of whether node i and node j can communicate on ch 1. This requires a channel switch ch 3 → ch 1 on node j as well as ensuring that node j does not interfere with the node for example 1 and 2 in ch 1. Similarly, it is ensured that node j can send with sufficient power to reach node i under interference from node 4 and 5 in ch 1.

The above procedure is executed until the paths and channels do not change. Then, when the new connection is established, the data can start to flow. After some time when there is no need for connection, it is removed. It is possible to support low stipulated mobility that the lifetime of the connections is relatively small in relation to the mobility of the node and / or the variations of the channel.

The algorithm works even when dynamic power adjustment is not possible. In this case, the adaptation of the physical link is a selection between transmitting with a transmission power of the fixed link Pfij (ON) and not transmitting completely (OFF). Preferably the transmission power of the link is selected to be Pfij in

as long as Pfij is in the interval between, otherwise the transmission power of the link is set to zero.

To assist the reader in understanding the basic concepts of the invention as well as the CFPR algorithm, an example of the operation of the CFPR algorithm will now be described with reference to Figs. 7-13, which represent a network having 36 distributed nodes over a square area and using 14 time intervals (TS). Each node is represented as a circle without padding. The source node is indicated by a black star within the circle while the destination node is indicated by a gray star. Each node has an ID that is written just to the right of the node. The connections between nodes are shown with links in different gray scales, where the gray scale represents the number of TS. The TS number is also represented in parentheses along with the link, halfway between the nodes that interconnects the link.

Fig. 7 shows a tree originated at the source with the ID 5. This represents the phase of the configuration of the connection of a first flow when the CFPR algorithm has generated the preliminary connections consisting of adapted paths, channels and link parameters . In this particular implementation of the CFPR algorithm, the lowest TS number is always chosen if there are equally good time intervals there. This is why interval numbers are assigned in order of number from the source node.

Fig. 8 shows the selected path to the destination node with the ID 31. Hence, all other preliminary paths have been ruled out except the optimal CFPR between the destination source pair.

When a second flow and connection is to be established, one can intuitively see that the links are selected such that they are not detrimentally interfering with the existing connection and vice versa. In Fig. 9, the ID of node 7 is the source and the ID of node 30 is the destination.

When the second connection has been established, it is noted that TS 1, 13, 12, 2 and 3 have been reused. In addition, two intervals are used concurrently between node 22 and 23. Fig. 10 illustrates the channels and the resulting paths for the first and second flow.

image3

Finally, the establishment of a third flow and the associated connection is shown in Figs. 11 and 12.

It should be noted that Fig. 7-12 only show the establishment, but not the release of the connections. However, the latter is trivial and therefore left out.

In the previous example, γ M = 8dB, γ R = 5dB and γ W = 8dB. The resulting CDF CIR for all active receptors is shown in Fig. 13.

Although collision-free tunnels or resources from source to destination naturally make one think of switched circuit connections, it should be understood that virtual circuits can also exploit these collision-free resources. In this case, the capacity on a link is generally shared between multiple connections. This normally requires the use of a programmer on each node to give each connection its negotiated capacity.

Additional / optional matters

Comments on margins and channel behavior

In the case of line of sight (LOS), as may often be the case in a network at the top of a roof, the channel will be relatively stable. Therefore, the different γ margins can be relatively small. However, when the strength of the channel fluctuates on a larger time scale than the duration of the packet or the interpolation depth, the γ margins should be selected with sufficient margin to ensure that interference to the existing or proper connection will not cause greater problems.

Cost directionality

Note that it is allowed to permute the order of the index i and j of kij in the given equation, so that the cost from node i to node j is considered instead of the other way around. Hence, the cost from the source node is not determined, but rather the cost to the source node. In this case, the source node can be more appropriately targeted as the destination node. Since the preferred algorithm is heuristic, it does not always provide a path with the lowest possible cost metric. One way to handle this is to determine the path twice - once with the source as the origin and a second time with the destination as the origin. One then needs to use a metric that considers that the flow is directed from the source to the destination:

image 1

Complexity

To reduce complexity, a series of measures can be taken. First, a reasonable number of neighbors N (i) should be selected to ensure the reasonable degree of network connectivity. A value of 6-10 neighbors should be enough. The search region of a suitable connection path can be limited. One way to accomplish this is to look for some distance or jumps around the shortest path between the source and destination nodes. This requires that a shorter path be established prior to the search and those nodes in the vicinity of the path are informed that they belong to the search region. Point out that other search region choices are also possible. A sensible restriction in the search region includes consideration of those neighbors who are closest to the source. One way to determine that the nodes are closer to the source is to run an ordinary shorter path algorithm as a first step before the CFPR algorithm is applied. If the CFPR algorithm uses the transmission power as a metric, it makes sense to use a similar metric such as the accumulated path loss from the source. As indicated above, many terms can be discarded in the previous calculations when they assume values such as 0 or ∞.

Creation of Sensitive Routes

Not all routes generated by a heuristic algorithm have to be seen as sensible. For example, the need for high load channels can cause a path that goes along a large zigzag line. There are at least three mechanisms that handle this in part, if this is considered as a major problem. First, by exploiting load control, the reduction of the channel will be less likely to occur. This in turn provides routes that are closer to a shorter, stipulated path that uses the same metric as in the CFPR algorithm. A second method is provided by limiting the scope of the route search as described under the previous complexity section. One way to achieve this is to use the Bellman-Ford neighbors with less cost to the source.

5

10

fifteen

twenty

25

30

35

40

Four. Five

fifty

55

Maximum allowable transmission power based on the CIR limit

Rather than limiting the transmission power to a γ margin less than the power received for any receiving part of an existing link, an alternative condition is to limit the transmission power so that the CIR for any receiving part of an existing link is not less than a threshold of CIR ΓM. The maximum transmission power allowed is:

image 1

where each receiving part of an existing link or preliminary path experiences the level of interference I (m) and îx (m) is the expected interference at node x from the nodes along the preliminary path. This puts a lower limit on the level of the experienced CIR. The result is that traffic will be rejected rather than allowing the CIR level of existing links to fall below the CIR ΓM threshold.

CIR balancing

When the transmission power levels have been determined during the optimization, it may happen that, in practice, the actual CIR levels deviate anyway from the desired CIR levels. This can be compensated by balancing the conventional CIR, both distributed and centralized, of the transmission power levels in the network. In other words, once a new connection has been configured, it is possible to balance the transmission levels to obtain the desired CIR levels (or other QoS measurement) in the network.

Alternatively, in particular for the centralized case, the balancing of the CIR is used as an additional step of the CAC phase. In case the CIR balancing fails, the connection is rejected. Note that the CIR CDF in Fig. 13 will then be a function of the step. The advantage of this particular approach is an improved overall performance.

Algorithm extended to a larger horizon

As it is possible that the basic operation of CFPR determines a channel that is inadequate, so to speak down the road, an extension to the CFPR algorithm is described here. As an example, we assume that channels 1 + 2, 1 + 2, 1 are free for node k, j and i respectively, k and i are not neighbors while j is a neighbor of both k and i. If node j selected to use channel 1 from node k because it has a lower cost than channel 2, that would cause node j and i cannot create a link. Obviously, it would have been smarter to assign channel 2 between k and j, but using channel 1 from j to i.

The way in which this is handled here is to allow node i to determine the properties of the link (for example the channel and the parameters of the link) from k to j, restricted to the fact that a link can be created from j to i. Hence, node i searches for the lowest cost combination for two links at the same time. However, even if two compatible links have been determined, only the link closest to the source is maintained. In the next step, another node may decide to use the link between node j to i when it looks for the most promising link combination, but discards the link from i to itself. The exception to this rule of rejecting the furthest link from the source node is for the destination station, which determines the last two links but does not rule out either link.

In the basic CFPR algorithm, only one link is considered at a time. In this version of the CFPR algorithm, one rather takes into account two consecutive links at the same time. The concept of horizon is introduced here to indicate the extent to which the algorithm works. The basic CFPR has the horizon = 1, while the CFPR version in this section has a horizon = 2. The horizon can be extended to any larger value, however potentially with a tremendous increase in complexity when many nodes and channels are involved.

Support for higher data transmission speed

Since different applications may have different requirements in data transmission speed, it is important to provide some support for different data transmission speeds. Two methods can be used to vary the flow of data end to end. In the first method, the γ R margins together with γ M and γ W are selected to support a link mode comprised of a code rate and a signal constellation. Different and simplified transmission speeds can be handled by specifying different requirements of the CIR (requesting a CIR that corresponds to a certain data transmission rate). Typically the signal constellation can vary from 2-BPSK to 64-QAM. This is also preferably considered in the conditional setting of the transmission power for Pmin and Pmax. In the second method, multiple paths are established and used together to provide the desired data transmission rate. Combinations of the two methods can also be used.

Application Layer Integration

As previously mentioned, it is possible to use other protocol layers as well as more than three protocol layers in optimization. For example, the application layer can be included in the optimization, preferably in combination with the three lowest protocol layers. For example, the application layer can host an adaptive application, capable of operating at different data rates but with an associated application quality and compatible with the data rate used. Many video and voice based applications are good examples of adaptive applications that allow multiple data rates. More specifically, when a new connection configuration is attempted, the optimization of the objective function (or algorithm) is performed with respect to multiple data transmission speed requirements (given by the application layer). For example, several data transmission rates, as previously indicated, can be supported using a combination of multiple channels (eg multiple time slots) between the nodes, through link adaptation (various combinations of the signal constellation and error coding rates without return channel) or a combination of both. In the integrated optimization of said four layer functions, the feasibility of a range of permissible transmission speeds is evaluated under given restrictions. In an exemplary embodiment of the invention, at each optimization step, any desired but not feasible high data transmission speed is eliminated from the additional optimization steps.

The application layer can alternatively be used in an integrated optimization together with only two of the three lowest protocol layers.

Extended algorithm for adaptive antennas and MIMO

The algorithm can be extended to incorporate both MIMO communication and adaptive antennas. In the case of adaptive antennas, the parameters of the physical layer are selected, such as the weights of the antennas in reception and transmission antennas while minimizing the transmission power. This is limited to not disturbing the current traffic and ensures that the desired receiver has sufficient quality (signal to interference and noise ratio).

When a sufficient number of antennas are deployed (in the group of adaptive antennas) and high directivity can be achieved, interference will cease to be the limiting factor for the network. Instead, it is the resources of the channel that will limit the load that can be transported over the network. In the extreme case, that is, the interference can be completely neglected, other optimization criteria are adopted, trying to minimize the number of jumps. You can also add restrictions to balance the free resources in each node, to increase the probability of finding a free path in each case.

MIMO works in a similar way by selecting the link parameters including the MIMO weights of the transmitter and receiver. Parameter selection is limited to minimizing the transmission power of the link while also fulfilling a MIMO data stream of the desired link.

Aspects of implementation

In general, the optimization algorithm together with the corresponding connection admission control (CAC) procedure can be implemented as physical components, logical components, microprograms or any suitable combination thereof, using for example microprocessor technology, digital processing of the signal or ASIC technology (Integrated Circuits of Specific Applications). For example, the algorithm can be implemented as computer programs for execution by a computer system. Computer programs can be written in almost any type of computer language, such as C, C ++, Java or even specialized proprietary languages. Indeed, in an implementation based on computer programs, the algorithm is assigned in a computer program, which when executed by the computer system determines the connections and manages the admission control. Preferably, however, the CFPR algorithm and the corresponding CAC procedure are more or less implemented in physical components, using ASIC or other submicron circuit technology.

Fig. 14 is a schematic diagram of a network node in which a CFPR algorithm according to the invention is implemented. Only those components of the network node that are relevant to the invention are illustrated in Fig. 14. The network node 100 comprises a control system 110 and a general radio transmission / reception unit 120 having a band processing module base 121 as well as a radio frequency (RF) module

122. The control system 110 preferably comprises a connection admission control unit (CAC) 112 and a routing unit 114, as well as a database 116 for maintaining the network information. The routing unit 114 includes the functionality to route the traffic by means of a routing table 115. In this particular embodiment, a CFPR unit 113 is implemented in the CAC unit 112 to determine a set of connection parameters, if possible. The CFPR unit 113 retrieves the relevant information about the existing connections as well as the link chain itself of the preliminary connection from the database 116 and / or directly from the control signaling between nodes, and executes a CFPR algorithm with a suitable objective function . The CAC unit 112 is configured to make a CAC decision based on the results of the execution of the CFPR algorithm. If the unfeasible set of connection parameters can be determined by the CFPR 113 unit in view of the given QoS requirements, the CAC 112 unit rejects the connection configuration request. On the other hand, if the CFPR algorithm produces a set of feasible connection parameters, the required connection is established. This is normally accomplished by updating the routing table 115 in the routing unit 116 with the new connection parameters, and sending the connection parameters to the network nodes involved using "flooding", sending the extension tree, source routing or any other conventional mechanism. This concerns first of all a centralized implementation. Next, however, the aspects of implementation concerning a distributed implementation will be discussed.

CFPR on demand implementation

CFPR can be implemented in a distributed way. This can be done using the concept of routing on demand, as previously mentioned. Although routing on demand is known from some prior art schemes, there are several corrections to the traditional demand routing approach.

The first issue is that each RREQ (Resource Request) in the network leads not only to a list of nodes along a preliminary route, but also specific details in resource assignments. For example, each reception interval for a preliminary connection can be associated with a reception power level. In case the maximum permissible transmission power based on the CIR limit is used, the level of interference and noise can also be included.

The reason for transporting this information is to ensure that the assignments of the interval and the transmission power do not interfere with the resources allocated along a preliminary connection. Therefore, the transmission power levels of the resources in a preliminary connection are also distributed with the RREQ. The reason for this is that a node i should be able to determine if any node along the preliminary connection will interfere when node i is being received.

The list of the preliminary connection (node ID) and associated information such as the channels used, the potentially used transmission power, the potentially experienced reception power (and the interference experienced) can be limited to some fixed length. This list then acts in a FIFO manner when the list is full. The reason for this is that one does not want to spend resources by sending useless information. Obviously, a node in the preliminary list will be of low importance when it is sufficiently far from node i, and is not harmed by interference from node i or interference node i.

A field for the cost metric, which is not simply a jump metric, is also transported in the RREQ. A particular metric discussed above was based on the transmission power, giving the level of the accumulated transmission power along a route.

When the destination node receives the RREQ, it selects a lower cost path according to the distributed execution of the CFPR algorithm and responds with a RREP that is retransmitted along the selected path back to the source node. The RREP is preferably sent with sufficiently high power in a channel that is essentially free of collisions in a larger area such that the adjacent nodes accidentally hear the RREP information. The nodes that accidentally hear the RREP information subsequently update their resource allocation databases. The details of the protocol for RREQ and RREP are known from the prior art and therefore are no longer discussed.

Each control message such as RREQ and RREP also incorporates the transmission power used, such that the receiving node can determine the path losses. This is feasible, when the channel is more or less reciprocal in the sense of the average path gain. In the non-reciprocal case, other well-known methods to estimate path gain can be exploited.

The complexity reduction schemes described above can be used in conjunction with determining the route on demand.

Finally, the RREQ may indicate depending on the bandwidth requirement the desired margins γR together with γM and γW as well, or the desired link mode or both. Similarly, the RREP announces the relevant connection parameters such as the transmission power, the received power, the desired margins, the relevant time intervals (channels) and so on.

Reduction of complexity of operation on demand

The construction of a complete tree structure with multiple preliminary paths, ultimately involves unnecessary processing, only one path will be used. This section suggests another version of proper on-demand operation to mitigate unnecessary processing of redundant paths.

Suppose the source node has an approximate idea of where the destination node can be found. This could be given, for example, by a shorter proactive path protocol such as DSDV that is updated slowly, being known a priori (ie fixed nodes) or even on demand. A RREQ can now be sent to the destination along the shortest path or along a region that follows the shortest path. In doing so, the RREQ collects the link information of the existing connections along the shortest path. This information contains the same information required for the CFPR algorithm to calculate a connection later, such as the channels used, the transmission power used, the reception power experienced and so on. When the destination node receives the RREQ, it processes the information collected by the RREQ through the CFPR algorithm and its derivatives. Note that the RREQ may contain a request for bandwidth that cannot be supported by a single connection. In that case, the destination node can determine multiple connections for a flow. Subsequently, one (or more) RREP is sent back to reserve the resources. The RREP will then contain the IDs of the nodes, the channel to be used and the parameters of the link (for example the transmission power).

The advantage with this approach is that the calculation is performed only at the destination node and the flooding of the RREQs is limited. As the calculation is not performed during the sending of the RREQs, the RREQs will be sent quickly through the network. Another advantage is that the destination node can ensure loop immunity, execute the CFPR algorithm both forward and backward (as indicated in connection with the cost directionality), and implement arbitrary (vendor-specific) algorithms.

A disadvantage is that the information contained in the RREQs can become very large for many long routes. One way to solve this problem is to incorporate intermediate termination points between the source and the destination, for example each 20th node or so.

Of course, the tree-based approach previously described on a limited region with a limited set of nodes can also be used.

Extended algorithm for access to the non-spaced channel

The CFPR algorithm can be extended to incorporate channel access techniques, that is, they do not depend on dividing the medium into channels of equal size with predictable channel limits. Examples of such a channel access scheme is the 802.11 DCF protocol. Note that the current operation of DCF does not allow to allocate resources repeatedly in the future.

In this case, the cost kij may imply delay, possibly in combination with the transmission power. Each node attempts to find a transmission window between or after the current transmissions for a packet of predetermined size. Link speed can be adapted to compress data packet transmissions over time. Since maximum link speed is used whenever possible, the transmission power can be adapted (minimized) while ensuring that reception is a factor γ above interference plus noise. Fig. 15 illustrates how various link speeds are used on different links. This causes a different duration to send the packets, the so-called transmission duration delay. Indeed, the route, the delay of access to the medium and the parameters of the link layer (such as the speed of the link and / or transmission power) are determined so that the communication is substantially non-interfering with respect to the existing connections thus as it ensures the link chain itself of the preliminary connection.

By properly selecting the transmission request delay, and the link speed (which affects the transport delay) for each node, it is thus possible to minimize the total end-to-end transport delay from the source to the destination. Once the fastest link mode is used, the transmission power of the link can be reduced as much as possible, while the CIR requirement for the link mode is still met. This really assumes that both the delay and the transmission power can be combined in the objective function, preferably in a weighted manner. If it is more important to minimize the delay (or transmission power) in the network, the corresponding weight coefficient simply increases.

The embodiments described above are merely given as examples, and it should be understood that the present invention is not limited thereto. Further modifications, changes and improvements that retain the underlying basic principles disclosed and claimed herein are within the scope of the invention as defined by the appended claims.

REFERENCES

[1] S. Lee and A. T. Campbell, "LOGO: Support of in-band signaling for QoS in mobile ad hoc networks", in the Proceedings of the 5th International Workshop on Multimedia Mobile Communication, 1999.

[2] C. Perkins, E. M. Royer and S. R. Das, “Distance vector routing on demand ad hoc,” in the Internet Draft, draft-ietf-manet-aodv-06.txt, July 2000

[3] D. Johnson and D. Maltz, "Dynamic source routing in ad hoc wireless networks", in Mobile Computing, Publ. Kluwer Academic, 1996.

[4] S. Chen and K. Nahrstedt, "Distributed Quality of Service in ad hoc networks", JSAC of IEEE, SAC-17 (8), 1999.

[5] C. R. Lin, "Multimedia transport in multisalt wireless networks", in the IEEE Communications Record, vol. 145, No. 5, October 1998.

[6] R. Nelson and L. Kleinrock, "Spatial TDMA: A multi-jump protocol for access to the collision-free channel", Trans. Common. of the IEEE, vol. COM-33, No. 9, pages 934-944, September 1985.

[7] Lichun Bao and J.J. García-Luna-Aceves, “Collision Free Topology Dependent on Channel Access Programming”,

[8] Grönkvist, J., "Traffic Controlled Space TDMA in multi-hop radio networks", PIMRC 98, October 1998.

[9] Grönkvist, J., "Assignment Methods for Space TDMA Reuse" MobiCom 00, Boston, MA, August 2000.

[10] Y. -C. Hsu and T.-C. Tsai, “Bandwidth Routing in a Multi-packet Radio Environment”, in the Proceedings of the 3rd International Mobile Computing Workshop, March 1997.

[11] Y. -C. Hsu, T.-C. Tsai and Y. -D. Lin, “Routing of the QoS in Multisalt Radio Package Environment”, at the IEEE International Symposium on Computer Communications, Athens, Greece, June 1998.

[12] C. -R. Lin and J. -S. Liu, "QoS Routing in Wireless Ad Hoc Networks", in JSAC of IEEE, SAC17 (8), 1999.

[13] C. -R. Lin, "QoS Routing on Demand in Multisalt Mobile Networks", in the INFOCOM Proceedings of the IEEE, 2001.

[14] T. -W. Chen, J. -T. Sai and M. Gerla, "QoS Routing Performance in Multimedia Wireless Networks, Multisalto", in the ICUPC Proceedings of the IEEE, 1997.

[15] Y. -S. Chen et al., “Routing of the QoS Mutitractura, of State Liaison, On Demand in a Wireless Mobile Ad Hoc Network”, to appear in the IEEE 2002 European Wireless Network; February 26-28, Florence, Italy (previously available on the Internet).

[16] W. -H. Liao et al., "A Multipath QoS Routing Protocol in a Wireless Mobile Ad Hoc Network", International Conference on Network Connectivity (ICN) of IEEE, Vol. 2, pages 158-167, 2001.

[17] W. -H. Liao, Y. -C. Tseng and K. -P. Shih, "A TDMA Based Bandwidth Reservation Protocol for QoS Routing in a Wireless Mobile Ad Hoc Network", to appear on ICC, 2002 (previously available on the Internet).

[18] A. Michail, "Routing and Programming of Algorithms in Wireless Multisalt Networks of Limited Resources," Doctoral Thesis, University of Maryland, College Park, 2000.

[19] C. Zhu, “Control of Access to the Environment and Routing of the Quality of Service for Mobile Ad Hoc Networks”, Doctoral Thesis, University of Maryland, College Park, 2001.

[20] M. Sánchez, J. Zander, “Reuse of Adaptive Routing for Multi-packet Radio Networks”, distributed through the Industrial Cooperation Program organized by the Radio Communications Systems group of the Royal KTH Institute of Technology, Stockholm.

[21] International Patent Application with International Publication No. WO 02/05493 A2, January 17, 2002.

[22] International Patent Application with International Publication No. WO 01/50669 A1, July 12, 2001.

[23] U.S. Patent Application Publication No. US 2002/0049561 A1, April 25, 2002.

[24] U.S. Patent Application Publication No. US 2002/0054578 A1, May 9, 2002.

[25] U.S. Patent Application Publication No. US 2002/0075869 A1, June 20, 2002.

Claims (7)

1.-A method for configuring the connection in a multi-hop wireless communication network, said method comprising the steps of: - determine, for a requested connection between a source node and a destination node, a set of connection parameters that include the path, the channel and at least one physical link parameter by expanding a directed tree with the preliminary paths for the originating pending connection at the source node and performing a search procedure to find at least the Ki cost for each node i, in a given set, from the source node according to the following nested algorithm: image 1   KID Source = constant, where i ≠ Source ID, N (i) is the set of current neighbors of node i which in turn is a set of all Ω nodes in the network, j is a neighboring node that belongs to N (i), m is a set of at least one channel in a set of M orthogonal channels in total, Ψ is one or a multitude of physical layer parameters, ki (j, m, Ψ), also indicated kij (m, Ψ), is the cost from node j to node i, in which the cost kij (m, Ψ) includes the transmission power of the link Pj (m) for node j and channel m as a parameter of the physical layer Ψ, and the power of Link transmission Pj (m) is subject to restrictions that restrict the transmission power of the link for a predetermined interval, and the term K (j), also indicated Kj, is the cost accrued from the source node to node j, and KID Source is the initial cost in the source node, at which the innermost level of nesting of said nested algorithm tunes to said parameter (s) of the physical layer Ψ, the next level of nesting is a choice of a set of channel (s) m for each neighbor, and the third level of nesting provides a choice between neighbors j, hence choosing the path in the routing layer; Y - establish the required connection based on the determined set of connection parameters. 2. The method according to claim 1, wherein said at least one physical link parameter is selected from the group of:

-

 the transmission power;

-

 the parameters of the adaptive antenna (AA);

-

 the multiple input parameters multiple output (MIMO);

-

 the modulation parameters;

-

 the bandwidth;

-

 the speed of data transmission; Y

-

 Error correction parameters.

3. The method according to claim 1, wherein a local search procedure is executed in each node i to evaluate the cost kij (m, Ψ) from node j to node i for all nodes j in said node set N (i) of neighbors, and the lowest cost Ki for each node i from the source node is determined based on said evaluation along with the information on Kj received from each node j. 4. The method according to claim 1, wherein the cost kij (m, Ψ) includes a channel that depends on the physical link parameter Ψ (m). 5. The method according to claim 1, wherein the transmission power of the maximum allowed and minimum link required from node j is defined as: image2 image3 indicate the transmission and reception power, respectively, for node r within the set R denoting the nodes along the preliminary path connected to node j, Pv (m) and Cu (m) on the other hand indicate the power of transmission and reception, respectively, for nodes with established traffic, γ M is a mitigation factor, G (m) is a gain matrix of the respective channel, γ R is a reception factor, and W 5 is a noise level that uses a γ W factor to ensure that it is generally limited by interference rather than limited by noise. 6. The method according to claim 1, wherein the transmission power of the maximum allowed and minimum required link from node j is defined as: 10 image2 image3 indicate the estimated (or rather preliminary) transmission and reception power, respectively, for node r within the set R indicating that the nodes along the preliminary path 15 connected to node j, Pv (m) and Cu (m ) on the other hand they indicate the transmission and reception power, respectively, for nodes with established traffic, G (m) is a gain matrix of the respective channel, ΓM is a threshold of the CIR (Carrier to Interference Ratio), which assumes that each receiver part of an existing link or path image2 preliminary experiences the level of interference J (m), is the expected interference at node x from the nodes along the preliminary path, γ R is a reception factor, W is a noise level that uses a factor γ W 20 to ensure that this is generally limited by interference rather than limited by noise. 7. The method according to claim 1, further comprising the step of determining a shorter preliminary path prior to said search procedure, and selecting said search region of neighboring nodes N (i) based on the knowledge of said Shortest preliminary path. The method according to claim 1, wherein the routing on demand is deployed, which has a Resource Request (RREQ) that transports the path, the channel and the physical link parameters for a requested preliminary connection and that it has a Route Response (RREP) that confirms a selected connection along a corresponding path. The method according to claim 1, wherein the routing on demand is deployed, which has a Resource Request (RREQ) that collates the information on the existing connections when said RREQ is sent through the network , and determining, in a destination node, a new connection that meets a Quality of Service (QoS) requirement as contained in said RREQ based on said collated information. 35 10.- A control system (110) for configuring the connection in a multi-hop wireless communication network, said control system comprising: - means (113) to determine, for a requested connection between a source node and a destination node, a 40 set of connection parameters that include the path, the channel and at least one physical link parameter expanding a tree directed with the preliminary paths for the pending connection originating from the source node and executing a search procedure to find a lower cost Ki for each node i, 45  KID Source = constant, image4  where i ≠ Source ID, N (i) is a set of current neighbors of node i which in turn is a set of all Ω nodes in the network, j is a neighboring node that belongs to N (i), m is a set of at least one channel in a set of M orthogonal channels in total, Ψ is one or a multitude of physical layer parameters, ki (j, m, Ψ), 5 also indicated kij (m, Ψ), is the cost from node j to node i, in which the cost kij (m, Ψ) includes the transmission power of the link Pj (m) for node j and channel m as a parameter of the physical layer Ψ, and the power Link transmission Pj (m) is subject to restrictions that restrict the transmission power of the link to a predetermined interval, and the term K (j), also indicated Kj, is the cost accrued from the source node to node j, and Source KID is the initial cost at the source node, at which the level of 10 more internal nesting of said nested algorithm tunes to said physical layer parameter (s) Ψ, the next level of nesting is a choice of a set of channel (s) m for each neighbor, and the third level of nesting provides a choice between neighbors j, hence choosing the path in the routing layer; Y - the means to establish the required connection based on the determined set of connection parameters.