FR2689657A1 - Radio communication system serving multiple stations - Google Patents

Abstract

Each of a number of network data processing stations is provided with a network interface (RE) interacting with a transmission/reception management device (ARC) which issues orders confirming transmission or reception.Each station comprises a multi-directional radio transceiver with a central channel transceiver (RD) for messages, an auxiliary transceiver (RI) changing in response to collision detection, and another transceiver transmitting an engaged signal as long as a signal is being transmitted over the main channel.

Description

Data transmission installation, of the radio network type, and corresponding method
The invention relates to computer networks which allow the exchange of information or "data" between different stations.

Such a network conventionally comprises a transmission medium, generally an electric or optical fiber cable.

Stations or stations are connected at various points on this cable, this connection being carried out through a "network interface".

It is necessary that the stations understand each other, despite their multiplicity In certain computer networks, a rule is established for this purpose (multiplexing, token) according to which at most one of the stations can transmit on the communication medium at a given time. Another technique allows "random access" to the network, that is to say that several stations can transmit at the same time. A consequence of this multiple access is that "collisions" can occur, and they will have to be resolved.

The ETHERNET network (registered trademark), governed by the so-called standard
IEEE 802.3, is of the random access type. The network management protocol is of the carrier sensitive multiple access type with collision detection, or CSMA / CD (Carrier
Sense Multiple Access with Collision Detection).

The implementation of such a network, but with ether as a communication medium, that is to say with transmission by radio waves, poses various problems that the present invention helps to solve.

A first object of the invention is to improve the detection or recognition of collision, in the case of a radio network.

The invention also aims to ensure that all of the stations or stations concerned will have joint collision information, which is necessary for the consistency of the collision resolution operations which will follow.

The invention also aims to ensure the proper maintenance of frequency synchronizations between the various stations connected by radio.

The invention also aims to take into account the particularities of radio transmissions, which mean that several routes exist, on a set of three non-aligned stations, but that obstacles may prevent some of these routes.

The invention also aims to allow indirect transmission of information between two stations, via one or more intermediate stations.

The data transmission installation proposed here is of the type comprising at least two data processing stations (the word "processing" being taken in its most basic sense). Each station is equipped with a network interface, capable of transmitting messages on request, and of collecting received messages. This includes a transmission-reception management device comprising a collision detector and a collision resolution means, following a recognized collision condition. One of the essential functions of this management device is to give confirming transmission or reception orders to the network interface.

According to a first aspect of the invention, each of the stations includes a multidirectional radio transmission-reception device (involving several directions, but not necessarily omnidirectional), with
a transceiver operating on a so-called cen channel
tral, for the transmission of messages to be exchanged, and
ordered in accordance with said confirmatory orders, and
an auxiliary transceiver operating on at least
a lateral channel.

In response to at least the start of a confirming transmission order, the auxiliary transceiver, capable of rapid transmission-reception switching, operates in rapidly alternating transmission-reception, according to a chosen pattern of its own (preferably substantially to be specific to this transmission-reception device.

In transmission, a collision is then recognized by the collision detector if a signal of level greater than a first level threshold is received during a substantial part of at least one silence of this side channel, manifesting the concomitant emission of a another post on this side channel.

In addition to this, the network interface is arranged to recognize a collision on reception by violation of a predefined criterion, associated with error detection and / or correction codes.

Preliminary collision information is thus defined, in transmission and / or in reception, at each station or station. It must also be ensured that the different stations will have the same information on the collision.

To this end, according to an important aspect of the invention, when it has recognized a collision, a station proceeds to a transmission on collision "or emission of collision signaling", and which consists in transmitting on a side channel a signal of determined time and frequency characteristics, for example a pure frequency of minimum duration.

Otherwise, it listens to the side channel to determine if another station is reporting a collision.

If, at the start, the station was itself "active" (ie at the beginning or "attempt" of transmission), it can proceed to an "active listening" consisting in accompanying for a predetermined duration a possible signal representative of the collision it would receive. In this way, any "active" transmitter propagates the "collision" information to all the stations which are within reception range.

According to another aspect of the invention, a second lateral channel is provided on which any station which receives a gain signal sufficient to be able to be processed ("demodulable" signal) will send a busy signal, which will be interpreted by all other station (within radio range) such as a broadcast ban.

According to another aspect of the invention, in the absence of sufficient natural traffic, the management device is arranged to force the transmission of artificial messages, in order to allow the mutual setting of pilot frequencies of the various stations to be maintained, as will be described later. This decision to send artificial messages advantageously obeys substantially random criteria.

Artificial messages can be used for service purposes (i.e. usefully, but not for the data to be transmitted itself).

On the one hand, they can be used to exchange information useful for understanding messages, in the event that they are transcoded. They can also be used to transmit information or tables defining relay modes between different stations, when it is a transmission between distant stations requiring intermediate relays.

Other characteristics and advantages of the invention will appear on examining the detailed description below, and the appended drawings, in which - Figure 1 is a very simplified diagram of a conventional computer network where the transmission medium is wired; - Figure 2 is a very simplified diagram illustrating four stations interconnected by radio transmission; - Figure 3 is the block diagram of the entire network interface, for a station according to the invention; - Figure 3A is a partial detail of Figure 3; - Figures 4A to 4D illustrate the format of the frames; - Figure 5 is a block diagram of a mechanism applicable to the detection of collisions; - Figure 6 is a more detailed block diagram of part of Figure 5, while Figure 6A is an equivalent electrical diagram, corresponding to Figure 6 and Figures 6B and 6C show a variant; - Figures 7, 8 and 9 are other more detailed block diagrams, relating to parts of the mechanism of Figure 5; - Figure 10 is an electrical diagram of the operations carried out at the second lateral channel; FIG. 11A is an equivalent electrical diagram of the operations of the collision detector automaton, while FIG. 11B is an equivalent electrical diagram of the other parts of the transmission / reception management device of a station; - Figures 12 and 13 are mechanisms allowing, by the transmission of artificial messages, to carry out frequency synchronization between the different stations; and - Figure 14 is a variant of Figure 5.

The accompanying drawings are, for the most part, certain. Consequently, they form an integral part of the description and can not only serve to supplement it, but also contribute to the definition of the invention if necessary.

Furthermore, the Applicant is filing today another patent application entitled "Data transmission device for random access network, with improved collision resolution, and corresponding method". Due to technical relationships existing with this Request, the description and drawings of this other Request can also be incorporated into this description.

In FIG. 1, a transmission medium MT is connected to network interfaces Ia to Ic, respectively connected to stations Pa to Pc. This is the classic structure of a data transmission computer network, to which the CSMA / CD protocol already mentioned can be applied, in accordance with the IEEE 802.3 standard.

The specialist knows that the models for defining such interfaces are defined "in layers" (ISO standard), to each of which is assigned a specific function. This division into functional layers makes it possible to ensure the compatibility of organs of the network from different sources, when they are interconnected. Reference will be made in the following to this notion of layer. A distinction will be made between the layers of the protocol proper, and the upper layers of the protocol.

The stations such as Pa are data processing stations (the word processing is used here in the most elementary sense, this processing can be very simple). All the processing carried out in the computer is outside the actual data transmission. However, there may be specific operations in the post which take specific account of the nature and certain conditions applicable to the transmission of data. These are the upper layers of the protocol.

The layers of the protocol itself will, on the contrary, govern the transmission of data in its basic conditions, in a way that makes it possible to ensure its security.

In FIG. 2, the stations or stations Pa to Pd are now connected to radio devices Ra to Rd, provided with antennas. In the simplified example illustrated, all stations can exchange data directly, except stations
Rb and Rd, between which there is an OBS obstacle.

What is called network-interface here, in a broader sense than for computer cable networks, is all that is between the station itself and the antenna (s) it contains.

The word "message" here designates a set of data to be transmitted, of any size. The word "frame" or "packet" designates the elementary data transmission unit, ie the block of data that can be transmitted together.

One of the features of the present invention is to ensure that when a station wishes to transmit a frame, this station is the only one to do so within its radio range.

Reference is now made to FIG. 3, which illustrates the general structure of a network interface for a station.

On the right, the RE block designates the standardized connection to the IEEE 802 standard, which allows connection to a network output of a computer. This connection is for example made with the integrated circuit i82586 sold by INTEL Corporation.

On its left hand, the RL block is a "local router" whose role will be explained below. Its function is to establish messages to be transmitted, with transmission requests, as well as to collect the messages properly received, and possibly to re-direct them.

Still on the left appears an ARC block, which is the transmission-reception management unit, in particular according to collisions
An ADF block is added to it, which will make it possible to correct frequency drifts, by sending artificial messages, when necessary.

To the left of the ARC block, there is at the top a main channel, which is established between a transmitting-receiving antenna A0, a modulated radio transceiver, for data transmission, denoted RD, and a CC circuit which has the function, on transmission, of embellishing the message with correcting codes, and, on reception, of verifying the correction of these codes to detect possible errors within a message (otherwise received normally, i.e. without collision).

To the left and in the middle is a first lateral channel, provided with a transmit-receive antenna A1, a transceiver R1 with fast transmit / receive switching, and an RC circuit for radio detection of collisions.

There is also a block A which operates as a generator of pseudo-random codes, such codes being used for different purposes according to the invention, on the one hand by the RC circuit, on the other hand by the block ARC already cited. The RC and ARC blocks are directly interconnected.

Finally, on the left and bottom, a second lateral channel is represented with its antenna A2, its receiver R2, and a circuit
RP, which allows radio detection of "hidden carriers".

The functional breakdown which has been carried out in FIG. 3 will not always be respected in the following, in order to facilitate understanding.

Thus, the ARC block can be broken down into three ARCA blocks,
ARCB and ARCC, as shown in Figure 3A. The ARCA block is responsible for directing the detection of collisions, and for exploiting the result. The ARCB block is used for the resolution of collisions, and the establishment of confirmatory orders for transmission or reception of data. A collision resolution mechanism can be incorporated into this ARCB block, which can be of any known type. Finally, the ARCC block will be responsible, in conjunction with the ADF block, for ensuring the transmission of artificial messages.

Reference is now made to FIG. 4.

FIG. 4A illustrates the conventional format of an ETHER message
NET compliant with IEEE 802.3 standard.

The present invention provides that this message is supplemented, if desired, by the additional indications which appear on the right in FIG. 4B. They include on the one hand an indication of relay recipient and relay source, a code called LARA code (this name "LARA" being assigned to the new protocol according to the present invention), and a buffer zone which can contain additional information.

When several relays are provided, the buffer zone contains, as shown in FIG. 4C, first of all a code indicating that it is a point-to-point transmission, that is to say a single transmitting station to a single receiving station, on the other hand an area indicating the number of hops already made by the frame in the various relays which it has undergone from the initial transmitting station. This buffer is updated at each jump or relay in the frame.

Conversely, a message can be sent by a station to several stations, what will be called a "broadcast". In this case, as illustrated in FIG. 4D, the buffer zone is multiplied, to include a code indicating that it is a broadcast, the number of relays, as well as the recipient addresses of different relays 1 to n , a broadcast message number, and the number of hops made by the frame in its previous relays.

What has just been described is the structure of a "frame" of messages. In a simple example, each frame consists of two parts, namely the header, and the body of the packet.

The header contains at least the address of the recipient (s), and the address of the sender, as well as some specific information about the traffic (for example whether or not an acknowledgment of receipt is required). The packet body contains the data to be transmitted (file, acknowledgment), as well as the end of packet signals.

These frames or packets are routed one by one from the RE block in FIG. 3. The RL block will process these packets, essentially in the case where a recipient is not in the radio range area of the transmitter; it then proceeds to modify the header by adding the address of an intermediate station, as illustrated in FIG. 4B.

The packets are then taken care of by the ARC block, which decides on the transmissions, and interrupts them in the event of a detected collision. The CC encoder and the RD device are responsible for the actual transmission of the frames in ether. When a frame has been correctly transmitted without collision, a signal is transmitted to the RE block, to allow the next packet to be taken.

On reception, the operation occurs in reverse. The CC block will check the correction of received packets; As a reference, packets with errors will simply be destroyed.

The RL block determines whether the message has arrived at its destination, or slil is being matched or relayed. A message arrived at destination is transmitted to the RE block, while a message to be relayed possibly sees its header modified taking into account the relay, to return to the ARC block for re-transmission.

The international standard ISO - IEEE 802 provides for a division into layers of the communication functions. The functions located from the antennas to the included ADF block constitute an adaptation to the radio transmission of the functions conventionally called "Medium Access Control" layer or MAC. In accordance with the present invention, a new layer is then introduced, which corresponds to the RL block, and which can be called the "membership and routing" or MR layer.

At the RE block, we find the interface layer
Classic ETHERNET.

We will now discuss the so-called MAC layer, which will be described in detail.

First of all, a precise method of data transmission must be applied at the level of the RD block. It is indeed a question of transforming a binary digital sequence into an electromagnetic signal suitable for radio propagation via the antenna, and vice versa, which is generally more delicate.

The aim is to provide maximum electromagnetic protection to the digital sequence which must pass between a transmitter and a receiver; it is particularly necessary to fight as effectively as possible against the phenomena of interference or multiple paths. On the other hand, it is useless to fight against interference by the possible emissions of the other stations of the network, because it is there the criterion which will make it possible to establish the existence of a collision, and then to resolve it.

Thus, the RD block can work by spectrum spreading with a unique code for all the stations of the same network, this unique code being able to be a pseudo-random code common to all the stations, and detected by filter adapted to reception. This is just one example.

It will also be noted that it is necessary to ensure suitable synchronization on the frequency level between the different stations involved. Those skilled in the art know how to ensure this synchronization in the event of sufficient traffic. When this condition is not verified, the drift between the different pilot oscillators can become such that the synchronism can no longer be caught very quickly. A solution to this problem will be proposed later.

Generally, a radio reception will be considered demodulable if its gain on reception is greater than or equal to a threshold value ed. If the gain is lower, the received signal is considered "received with errors". Advantageously, a value Ed is defined, which corresponds to reception "in the immediate vicinity" (This is used in particular for routing, described below). This definition of a threshold common to all stations or stations assumes that the reception power is the same for all, which is generally accepted here.

The function of the DC circuit, in transmission mode, is to transform the IEEE 802 standard frames into binary sequences more resistant to transmission errors, thanks in particular to the use of error correcting codes (for example interleaving, associated use of a convolutional code).

In reception mode, the component restores the initial frame by correcting any transmission errors on the binary sequence received from the radio data reception device (when the latter has demodulable reception).

In addition, it detects any residual errors when processing the reconstructed frame, and signals these errors.

This may result in a "frame received with errors" state.

Preferably, two coding detectors and error correctors are used: - one for "dispersed" errors (VITERBI coding), - the other, dealing with block errors (REED coding
SOLOMON, with interlacing).

This supposes the implementation of said codings at the first transmission of the data (at least), and the implementation of control processing upon (or each) reception.

This or these codings can be supplemented by an interleaving technique where, after coding, the order of the bits is substantially modified in each sequence of bits sent; so a brief disturbance only affects bits of information that were very far apart from each other in the original sequence. The normal order is of course restored after reception before operating the error correcting decoding.

It is also possible to improve things further by taking into account physical data such as the instantaneous gain of radio reception, or else statistical considerations on the diversified antennas or the multiple propagation paths.

In the CC block, there is provided, upstream of its encoder part, an elementary device for rapid detection of errors, such as staking of the frame by parity bits in sufficient quantities. On reception, the error control operation (parity control for example) is performed downstream of its decoder part.

Reference is now made to FIG. 5, for the description, in the form of a method, of the implementation of the invention.

In FIG. 5, step 500 marks the wait for the start of a transmission (transmission) or a reception.

On such a start, step 501 tests whether it is a transmission or a reception.

As regards a transmission, we go to step 510, detailed in FIG. 6; contrary to reception, we go to step 550, detailed in FIG. 7.

In either case, the following steps 520 respectively 560 determine whether a collision condition has been recognized, which is materialized by logic signals
COLL11 and COLL12, respectively (0 = false; 1 = true).

If it is a collision, go to steps 522 and 562 respectively.

In the absence of a collision, we go to step 530 of active listening if it was at the start of a program, and on the contrary to step 570 of passive listening if it was a question of 'a reception. These steps 530 and 570 are detailed respectively in FIGS. 9 and 8. As a variant, passive listening is carried out in both cases.

Then, we go to steps 540 and 580 respectively, to fix a collision state according to the result of the previous step.

In one case, the station was at the start of transmission. Step 541 determines the presence or absence of a collision. In the presence of a collision, the transmission is interrupted, preferably after insertion of a sequence provided with a parity error, on special command addressed to the CC block, as shown in step 542. In the absence of collision, the transmission continues as indicated in step 543.

In the other case, it was initially a reception. In the absence of a collision, it continues at step 585 with however a rejection of the current frame if it contains a parity (or other) error.

The detail of step 550 is relatively simple, and will now be described with reference to FIG. 7. At the start, sub-step 551 establishes a false collision condition (COLL12 = 0). In step 552, the block CC searches for a possible parity (or other) error in the incoming frame, and this for a fixed duration. In sub-step 553, the signal COLL12 is passed to 1 if such an error is detected.

More complex, the detail of step 510 will now be described with reference to FIG. 6.

The pseudo-random generator GA (FIG. 3) supplied a pseudo-random word Gc having Nc bits.

In step 511, a processing index i is initialized to 1, while the variable COLL11 is set to the value 0, representative of "false" collision.

Step 512 consists in taking from the word gc its ith bit, which will be designated by a.

Step 513 tests whether the bit a is 0 or 1. If it is 1, step 515 consists in transmitting on the first side channel for a time Lc. If it is 0, we will on the contrary go to step 514 to listen on the first side channel during the time Lc.

In step 516, it is tested whether a gain carrier greater than a threshold value ec has been detected. If yes, it is because another station has transmitted at the same time (on the basis of a different pseudo-random word). In this case, step 517 consists in passing the variable COLL11 to 1.

In either case, we then return to test 518 which determines whether we have reached the last bit of the pseudo-random word GC. If this is not the case, step 519 increments the work index i, and we return to 512.

The final output is to go to the next collision reporting step, which is step 520 in Figure 5.

Advantageously, the carrier detection of step 516 is not carried out immediately from the start of the listening time interval defined in step 514. A time lc is preferably expected, so that the collision detector cannot be abused by the echo of its own broadcasts.

Furthermore, it is ensured that the transmission of step 515 appears continuously, when two consecutive bits of the pseudo-random word gc are at one.

What has just been described relates to the first step in the operation of the collision detector automaton, a step which can be carried out either in an active state or in a passive state, depending on the result of test 501 of FIG. 5.

The second stage follows the first, after a silence. This silence is preferably at least equal to Lc. In all, the second step therefore begins (Nc + l) * Lc units of time after the start of the action of the collision detector automaton.

During the second step, this automaton has at least two states, namely a transmission state to signal or "broadcast" the collision and a listening state. It preferably has three states, the listening state can be passive listening or active listening.

We now consider two new reference quantities which are a threshold of level Ec at least equal to the threshold ec already invoked. We also consider a new time constant Jct larger than the time Lc.

When a collision has been detected (COLL11 = 1, respectively COLL12 = 1) in the local station, the global steps 522, respectively 562, of FIG. 5 consist in transmitting on the first side channel a predetermined signal, such as a pure carrier, for Jc time units.

The passive listening of the global step 570 is detailed in FIG. 8: the sub-step 571 consists in listening on the first side channel during Jc time units; sub-step 572 searches on this channel for a carrier having a gain greater than Ec, for a substantial duration, at least equal to Lc; if such a carrier is detected during the time thus defined, step 573 validates a collision variable
COLL22 to 1.

In a way, each station defines its own (irregular) transmission / reception comb. Lc is the clock used to define this comb. Lc must be chosen greater than the time during which we receive echoes after transmission (fading). Jc is a time (in principle predetermined), chosen sufficient so that, in the presence of a collision, all the stations concerned can be aware of it.

If the station was in transmission previously, but did not detect a current collisio at the output of the register RC60 and its complement established by the inverter RC611. If the bit is 1, the transmitter R1E is energized. If the bit is 0, and after a delay defined by a stage RC616, the output of the receiver R1R is monitored, a logic signal COLL11 being defined by an AND gate RC618. If a transmission is obtained while the current bit of the word GC is at 1, and after the delay 1c already mentioned, COLL11 is true (COLL11 = 1 in memory RC619).

During the second step of the PLC, the signal S21 is true, in the event of a collision already detected locally, and the transmitter R1E is energized via the OR gate RC625. If it is false, after complementation by the inverter RC630, the output of the receiver R1R is monitored, by an AND gate RC635, which determines a collision condition COLL21, stored in memory RC639, and corresponding to listening (passive, in this figure 6A).

It is then a collision of which the station is informed by another station, without having detected it itself. The sequence may be purely passive or active, depending on predefined criteria, which may be other than those set out above, and depend in particular on the chosen collision resolution mode.

If it is necessary (active listening) to accompany the program so that it lasts for a well-defined time, it suffices, for example, for the duration of the COLL21 signal, to directly command the transmitter R1E to carry out the extension broadcast desired in the first lateral channel, on the pure carrier of "signaling" or "broadcasting" of the existence of a collision.

FIGS. 6B and 6C illustrate a variant in which the first lateral channel has been split into two "first" separate lateral channels, suffixed 1 and 2 respectively in their transmission and reception parts. This results in simplifications of circuits. FIG. 6B relates to the first step of the automaton, for the case of a station having transmitted. FIG. 6C relates to the second step of the automaton, where we have distinguished (simply for illustration) the case S21 of the stations which have already recognized the collision and that S22 of the stations which have not yet recognized it. The elements of the two channels RC635,645 then RC 639,649 are the same. And an OR RC660 function combines the COLL21 and COLL22 outputs with
COLL2.

The operation is as follows:
The two steps S1 and 52 can take place simultaneously: you can listen to R1R2 during the first step, which takes place as before with RlE1 and R1R1 for an active station. Any collision detected forces immediate emission to R1E2.

The functions pertaining to the second (or third, as the case may be) lateral channel will now be described with reference to FIG. 10. As previously (FIG. 6A), the antenna A2 and the radio stage R2 are subdivided into transmission parts A2E,
R2E and receiving parts A2R, R2R. Similarly, for the main channel, its AOE and RDE reception part is separated, and its AOR and RDR transmission part. In the CC block, a distinction is also made between its CCE transmission part and its CCR reception part.

In the event of reception, the CCR coder supplies data on the one hand, and on the other hand an indication RE that it is in reception (of signals demodulable by the RDR stage). The signal RE forces the emission of a busy signal by the stage R2E of the second lateral channel, signifying that the ether is occupied, and this for all the stations within radio range.

Conversely, apart from this case of demodulable radio reception, the R2R reception part of the second side channel is arranged to detect the existence of such a busy signal transmitted by another station (above a threshold level), and then provide a logic signal BP = 1, indicating an occupation of the ether on the main channel.

A logic function ET LL200 receives the basic transmission order TR, and the signal BT supplemented (the complementation is shown by a small circle on the input of the door
LL100). The basic transmission order TR therefore only becomes a confirmation transmission order TRE if no busy signal
BT is not received. If TRE is true, the transmitter stage RDE is authorized to transmit packets or data frames, which it receives after suitable transformation of the transmission encoder circuit CCE.

The busy signal can be used in several ways: - in the embodiment described, it is sent by all stations receiving a demodulable signal - a first variant consists in interrupting this signal as soon as a reception error is observed; the absence of a busy signal can then serve as a collision indicator, possibly even being the only one, under certain conditions; - in a second variant, only the destination station (s) (primary) of the message emit the busy signal (as desired, all or none emitted at the start, during the time necessary for decoding the recipient) - another variant consists in transmitting a busy signal when the received signal is not only demodulable but also decodable. In this case, the collision can be defined by the fact that in the presence of a transmission, no busy signal is emitted.

Some of these variants can be combined.

Reference is now made to FIG. 11A, which describes the ARCA part of the ARC block of FIG. 3.

An ARCA10 sequencer defines steps 1 and 2, for example as a function of a logic start signal DEB and a logic signal TR / RE indicating whether it is a transmission or a reception.

This results in logic signals S1 and S2 representing respectively the first and second stages of the collision detector automaton. If one is in transmission (TR), a logic gate LL300 puts in true a signal S11 (first step, active state). This signal S11 controls the circuits of FIG. 6A already described. These can provide in return a collision signal COLL11, detected on a local broadcast.

If we are receiving, an ET LL310 door intervenes. It has an input receiving the signal RE representative of reception, an input receiving the signal S1 and a third input receiving, complemented, the signal REOK.

Reception of an incorrect frame will establish a detected collision logic signal COLL12, representative of a reception collision, attributable to other stations.

An AND gate LL320 combines the signals COLL11 and COLL12 into a signal COLL1, stored in a memory LM325 (local collision detection). If a collision is stored in
COLL1, a gate LL30 brings together the signals COLL1 and S2 to restore in the second step the state of emission on collision, by a logic signal 521 = 1.

Otherwise (521 = 0), active or passive listening will provide a collision signal COLL22 = 1 if necessary. It indicates a collision recognized by other stations, but not detected by the local station. The joining by an OR function LL335 of the signals COLL1 and COLL22, stored in LM340, provides the definitive collision signal detected COLL.

The elements which have just been described can be the subject of different variants.

First of all, it is possible in certain cases to carry out collision detection only in reception mode.

In this case, the device managing the pseudo-random alternating transmissions / receptions on the first side channel can be eliminated.

Then the collision does not always occur at the start of the packet. An error can indeed be found during the packet ("late collision"). In this case, predefined criteria determine whether or not to report a collision: by simplifying it can be decided a priori to always report a late collision, or on the contrary never to report it.

FIG. 11B illustrates in a simplified manner the general functions of collision resolution.

The minimum required signals are - a transmission request signal in progress REQ ~ TR, - a reception signal in progress REC, - a busy signal of the data channel BT, and - the collision signal to be processed COLL.

All this is applied to an ARCB 20 automaton, which operates according to a known collision resolution method, such as BEB ("Binary Exponential Backoff" or Binary Exponential Shift). As a variant, it may be one of the methods recently introduced for achieving deterministic collision resolution, that is to say in limited time. Advantageously, this is the collision resolution method described in the aforementioned Patent Application.

Schematically, the ARCB block comprises a memory of local frames ARCB30, and a memory of radio frames ARCB40, with in between a switching mechanism ARCB50, which makes it possible to - process the signals received if one is in actual reception of usable signals , - decide if you have the right to transmit, in the absence of collisions, - resolve a collision, by determining when you will try to transmit, and by indicating when this attempt to transmit on collision will be considered successful, that is to say not itself generating a collision.

According to an interesting aspect of the invention, the switch
ARCB50 in FIG. 11B also receives signals from an additional memory ARCC60, which contains artificial messages to which we will return later. It is naturally associated with it an additional transmission request signal REQ ~ TR ~ SUP.

Before explaining the primary role of these artificial messages, we should briefly review the operation of radio transceiver devices.

These devices generally suffer from what is called frequency drift: oscillating devices - quartz and others - undergo frequency variations, which can go as far as compromising the efficiency of reception. Of course, it is known how to arrange such devices, so that when a reception device is shifted in frequency with respect to its transmitter, the receiver automatically adjusts to the second in an attachment phase. This attachment phase has a bearable duration when the initial offset is small. On the other hand, it could become extremely long if this difference is significant, and thereby compromise the functioning of a radio computer network.

The Applicant therefore considered that it is desirable, when the radio traffic is insufficient to ensure a sufficiently short time for synchronization of receptions on the broadcasts, to artificially increase this traffic by artificial messages.

A considerable difficulty then arises: the artificial traffic must ensure a permanent and not insignificant speed to ensure permanent registration of the various radio transceivers. But it must also be able to disappear very quickly when natural traffic becomes sufficient.

This must, in addition, be obtained by taking into account the particularities of the radio transmission medium, which is a distributed transmission medium, capable of many different configurations.

After analyzing this situation, the Applicant considered that it was necessary to ensure an appropriate statistical balance between transmission and reception: in other words, the sum of the flows of natural and artificial traffic must be equal to a quantity established at advance. However, this property must also be verified in each of the cells (range areas) surrounding each station or station. Those skilled in the art will then understand that the theoretical resolution of this problem involves calculations of enormous complexity, which it is out of the question to implement in real time and in situ without seriously compromising the operation of the network.

A compromise solution will therefore be proposed here, which appears to give satisfaction.

The ADF block is responsible for determining how to generate the transmission of artificial messages. This ADF block (figure 11B) is articulated on a chronometric circuit
ADF70, which defines control times with a TA periodicity (step 700, fig.12). For example, pulses of period TA are applied to an AND function ADF72, which also receives the local request signal for transmitted transmitted complement, REQ ~ TR / (the final slash marks a complemented signal). If, at time TA, there is no local transmission request (test 702), then we will close the switch ADF76 to authorize an ADF74 register to supply a pseudo-random number NPA, which will have been obtained (704) of the generator GA of FIG. 3. This pseudo-random number is applied to the negative input of a digital comparator
ADF88.

We denote by g (t) the current value of the pseudo-random number
Postcode.

In parallel, the set of blocks ADF82 and ADF84 maintains a probability quantity p (t). The quantity p (t) is transferred to the positive input of the digital comparator ADF88. If the comparator indicates (test 706) that g (t) is less than p (t), then we will produce the additional transmission request REQ TR SUP, to request the transmission of artificial messages (708).

If there is an effective data frame awaiting transmission in the ARCB circuit, we will simply inhibit the output of comparator ADF88.

The definition of the quantity p (t) is related to the observation of the medium or transmission channel, for which we are compelled to have a limiting number of different states or rather "observations", of which one and only one is valid at any moment. It can be the observations - collision slice, - empty slice, - transmission in progress (transmission or reception).

One way of defining these observations is described in the patent application of the same day already cited.

The quantity p (t) is defined repeatedly from its previous value and from the observation of a change of state of the channel.

At each observation of state (changes of state as well as new event of the same state - for example new collision while one is already in the state of collision), one updates p (t) according to its value previous and observed condition:
h [p (t), o] -> p (t) where "o" denotes the observed state.

Any function h [] which ensures that p (t) decreases in the event of an abundance of collision slices and increases in the event of an abundance of empty slices may be suitable.

With the three states defined above, and the collision resolution mode of the parallel request of the same day, the Applicant currently considers it preferable that: - h [p (t), "collision"] = MAX {fm, a * p (t)} (step 806) - h [p (t), "empty"] = MIN {Fm, b * p (t)} (step 804) - h [p (t), "transmission"] = p (t), as indicated by the direct return from step 702 at 700, where - a and b are two constants, with a less than 1 and b greater than 1, - fm and Fm are two other constants, with fm less than
Fm, itself less than or equal to 1, - MIN () and MAUX () are two functions respectively delivering the minimum and the maximum of two values.

It remains to define the content of artificial messages.

These can be purely artificial, i.e. only consist of zeros, for example.

They can be used to transmit the routing information to which we will return below.

They can also be used to transmit information useful for retrieving data, when these are transcoded, which can be done in different ways.

We will now describe the functions of the routing processing block RL of FIG. 3, which corresponds to the new layer called "MR", proposed in accordance with the invention. Reference is also made to Figures 4 for message formats.

It is recalled first of all that the standard IEEE 802 or RE connection (FIG. 3) recognizes, at the base, messages whose format is that of FIG. 4A.

As regards frames to be transmitted, the first function of the block
RL is therefore to examine the recipient address contained in the header of Figure 4A. The RL block then consults its local routing table. If necessary, it will modify the header of the frame, so that the latter is transmitted to an intermediate station which is within radio range. The format of the packet or frame is then that of FIG. 4B.

This done, the complete frame is sent to the ARC block, to process the transmission, resolving any collisions.

On reception, when a frame is received without error, the block
ARC transmits it to the RL block, which first determines whether the transit or relay address contained in the header is that of its local station. If this condition is not satisfied, the frame can be discarded.

If the transit recipient is the same as the actual recipient, it is a frame arrived at destination, and this is directly transmitted to the RE block, after correction of its header to pass from the format of FIG. 4B to format of Figure 4A.

In the opposite case, the router acts exactly as if the frame had been obtained from the RE connection, with transmission request: it consults its local routing table, and again modifies the header of the frame, to register it a new intermediate station, or the final recipient, after which the frame is returned to the ARC block. It is possible that a copy will be submitted to the RE block for information.

It remains to define the local routing table. This can be from an initialized or pre-established version, but it is desirable to update it constantly. This update can be carried out from the frames normally received from the ARC block.

Each frame normally received includes the indication of its immediate sender. It is therefore possible for the maintenance block RL to permanently define a table of neighboring stations or stations that it receives. In addition, at regular intervals, the RL block will create a spontaneous frame, by which it broadcasts to the other stations at its radio range at least part of the content of its routing table (stations it receives). This spontaneous frame can be incorporated into artificial traffic when it is created. It can be created by authority when there is no need for artificial traffic.

This means that each station knows for itself the stations it receives, and, by information transmitted, the stations which receive it. This results in the possibility of constructing two tables, which will be called - transmission register (the stations that we hear) - reception register (the stations from which we are heard).

For a given station (noted A below), the stations which are in the two registers are in relation to "symmetrical range" with this station. Routing can only take place using stations that check this symmetrical range condition.

From there, each station A can communicate directly (1 transmission) with its "symmetrical range" counterparts, indirectly (2 transmissions) with other stations with "symmetrical range" of the former, which serve as relays, and so after. It is conceivable to set a limit on the number of relay stations admitted, which will be called below "number of hops".

It is now possible to build a third table which will be called "topological directory" for each station. This directory is a rectangular matrix or with two entries with for example in columns the numbers of the stations registered in the emission register of A, that is to say which directly receive the current station, and in ordinates the numbers of all the stations which are accessible from the current station. Each cell in this table will contain the number of hops necessary to reach the target station finally.

Those skilled in the art will understand that the updating of this topological directory can be done from the routing tables received by the station considered from the other stations which are within radio range.

To carry out this update, it suffices to consider any decodable frame received "on the fly" by this station by examining in the header of its frames what is the intermediate source, what is the real source, and what is the number jumps already made by the data packet considered.

In fact, the message format buffer zone of FIG. 4B basically includes at least one code indicating that it is a point-to-point transmission, that is to say of interest only to a transmitting station initial and a final receiving station with the number of hops that it is planned to make to go from one to the other.

Again, it is convenient that each station transmits its chatter table at regular intervals, which allows verification of the consistency of the different routing tables of the different stations.

An interesting variant is "broadcast routing".

This is done on a special request, using a special form for the recipient address in the frame
IEEE 802.3 in Figure 4A, for example the fact that this address begins with a 0. This will mean that the message concerns a specific subset of the stations on the network, or even all the stations on the network.

It is then necessary to have two additional tables - a distribution table, and - a distribution directory.

The broadcasting table collects the numbers of stations with symmetrical range, or a subset of these (this subset having to check the property that there is always a station in the broadcasting table which makes it possible to reach any of the stations of the network, or any of the stations targeted by the broadcast).

The broadcast directory stores the broadcast numbers of a number of the latest broadcast communications that have reached the local station.

The Applicant considers that it is then advantageous to use the format of FIG. 4D, where the buffer zone comprises a broadcast code, a number of targeted relays, different recipients, a broadcast number and a number of admissible hops.

The invention can be modified in different ways.

For example, a side channel can be used to detect collisions, for example as described above with reference to the busy signal, but without resorting to collision reporting by a side channel.

Similarly, the use of the busy signal can be done without resorting to collision reporting by a side channel.

In addition, the words used here should be interpreted broadly. For example, "to carry out a transmission on collision via a lateral channel" covers the case where one would modify a transmission, in its carrier or any modulation of this one, and even the case where one stops the emission. Many such reversals are possible.

It is also possible to generalize the notion of lateral channel.

Thus, at least some of the side channels can be multiplexed "in frequencies". In other words, a side channel thus frequency-multiplexed will simultaneously occupy several separate subchannels, by transmitting on each of these subchannels a pattern, which is likely to differ from one subchannel to the other. 'other.

On the other hand, at least some of the side channels can be time multiplexed with the main data channel. A side channel thus multiplexed in time with the main channel will physically form only one channel at a time with it. In this way the history of the main channel takes place as a succession of pieces of side channel separated by pieces of data channel.

The parts of the lateral channel can be placed either before or after the packet to be transmitted, or even located in the body of the packet (which requires the interruption of the transmission of data in progress, and its possible resumption after insertion).

Finally, time and frequency multiplexing of the side channel or channels can be envisaged simultaneously.

We will now describe examples of such embodiments.

A first example is illustrated in Figure 14, which looks like Figure 5.

Before each packet transmission, as soon as the data channel is recognized free (500), a station which wants to transmit (501) inserts directly on the channel the collision detection step described by FIG. 5, namely: a phase 510 for collision search (using alternating transmission and reception) and a collision reporting phase 522. If a collision is detected and reported (540), the transmitting station gives up transmitting its packet (544). Otherwise, where no collision has been reported, the station transmits (545) its packet just after the end of the step described above. The receiving stations wait (551) for the duration of the collision search. They can signal if an error has occurred in the reception of the packet (passive detection 570). Alternatively, the collision report can be made on a physically separate side channel. Then the received packet is subject to a parity check (586).

In a second example, it is assumed that the data channel is already frequency multiplexed, and that the transmission of each packet occupies part or all of the subchannels. We can then consider the following installation Before each transmission of parquet, we place a header on one or more subchannels, which asks them to act each as a side channel part. Each of these subchannels is used for a short time to transmit a particular pattern as provided in the description of the side channel.

When a collision is detected by detecting a signal during a listening phase on at least one of the subchannels used, the collision is signaled, either on one or more subchannels of the data channel , or on a lateral channel.

Claims (25)

Claims 1. Data transmission installation, of the type comprising at least two data processing stations (Pa, Pb), each provided with a network interface (Ia, Ib), capable of transmitting messages on request, and of collecting received messages, as well as a transmission / reception management device (ARC) provided with a collision detector, and a collision resolution means, following a recognized collision condition, this management device giving confirming transmission or reception orders, characterized in that each of the two stations comprises a multi-directional radio transmission-reception device, with a transceiver (RD) operating on a so-called central channel, for the transmission of messages to be exchanged, controlled in accordance with said confirming orders, and an auxiliary transceiver (R1) operating on at least one lateral channel, having a normal state, and in that in response to a detected collision, the the transmit / receive state of this side channel is changed. 2. Installation according to claim 1, characterized in that the network interface is arranged (RC) to recognize a collision in reception by violation of a predefined criterion. 3. Installation according to one of claims 1 and 2, characterized in that the auxiliary transceiver is capable of rapid transmission / reception switching, and in that, in response at least to the start of a confirmatory order of transmission, the auxiliary transmitter / receiver (R1) operates in rapidly alternating transmission / reception, according to a pattern chosen to be specific to this transmission / reception device, a collision being recognized by the collision detector if a signal of higher level at a first level threshold is received during a substantial part of at least one silence of this lateral channel, manifesting the concomitant emission of another station on the lateral channel. 4. Installation according to claim 3, characterized in that the chosen pattern is substantially random. 5. Installation according to one of claims 1 to 4, characterized in that the collision detector (ARCA) is an automaton, having a first step, where it can take a first state called passive and a second state said active, this automaton being put in the active state on a confirmation order of transmission, while it is otherwise in the passive state, in particular in response to the recognition of a beginning of reception of message, in that in sound active state, the collision detector automaton controls the auxiliary transmitter / receiver in rapidly alternating transmission / reception according to a chosen pattern, while in its passive state, it can place this auxiliary transmitter / receiver in reception, and in that the collision detector automaton has a second step, where it has a first state for emission on collision and at least one other state for listening, while any collision recognized during the first step forces the second step to the emission on collision, to inform other stations of the collision, and that otherwise the PLC listens. 6. Installation according to claim 5, characterized in that the first lateral channel is split into two first lateral channels, one of which is used for the functions of the first step, and the other for the emission on collision of the second step , which can be immediate. 7. Installation according to one of claims 5 and 6, characterized in that the collision detector automaton (ARCA) has in its second step, a second state called active listening and a third state called passive listening, one or the other state being taken according to the previous state during the first step, the active listening state ensuring an emission intervention on the lateral channel, to accompany any emission on collision, so that it has predetermined characteristics . 8. Installation according to one of claims 1 to 7, characterized in that each radio transceiver device operates on a second lateral channel (R2), on which it transmits a busy signal as long as it receives a demodulable signal on its main channel, while data transmission is prohibited for any station whose radio transceiver receives the busy signal on its second side channel. 9. Installation according to one of claims 1 to 8, characterized in that in the absence of sufficient natural traffic, the management device is able (ARCC; ADF) to force the transmission of artificial messages, in order to maintain a mutual setting of the pilot frequencies of the different stations. 10. Installation according to claim 9, characterized in that each management device proceeds to said transmission of artificial messages in a substantially random manner. 11. Installation according to one of claims 9 and 10, characterized in that at least some of the artificial messages are used for service purposes. 12. Installation according to claim 11, characterized in that at least some of the artificial messages are used for the exchange of information useful for understanding the messages. 13. Installation according to one of claims 11 and 12, characterized in that each management device (ARCB) is capable of controlling the transmission not only on local request, but also as a transmission relay between distant stations, and in that at least some of the artificial messages are used for the exchange of browsing tables, defining these transmission relays. 14. Installation according to claim 13, characterized in that each management device maintains at least a list of the stations which it receives, and transmits it as an artificial message, which allows it conversely to maintain a list of the stations which receive it from the artificial messages it receives. 15. Installation according to claim 14, characterized in that each management device also maintains  - a topological directory designating at least part of the stations with which it can communicate, directly or indirectly, and defining the route or routes of this communication, and  - a routing table defining at any time a preferential route. 16. Installation according to one of the preceding claims, characterized in that the messages are in standard format ISO for IEEE 802.3, plus a preliminary header. 17. Data transmission method, according to which - at least two data processing stations (Pa, Pb) are provided, each provided with a network interface (Ia, Lob), capable of cooperating with a common transmission medium to send messages on request, and to collect received messages, as well as to establish confirming send or receive orders, characterized in that at each station, the transmission / reception of data is carried out on a channel so-called central multidirectional radio (RD), controlled in accordance with said confirming orders, and in that, in response to a collision detection, it is carried out on a lateral channel (R1), a collision emission, of predetermined characteristics, for notify other stations of the collision. 18. The method of claim 17, characterized in that a reception collision is recognized by violation of a predefined criterion. 19. Method according to one of claims 17 to 18, characterized in that at least at the start of a confirmation order of transmission in a station, it is carried out on a side channel (R1), a transmission / reception rapidly alternated , according to a selected reason specific to this station, a collision being recognized if a signal of level higher than a first threshold of level is received during a substantial part of at least one silence of this lateral channel, manifesting the concomitant emission of another post on the lateral canal. 20. Method according to one of claims 17 to 19, characterized in that any station which receives demodulable messages on its main channel transmits on a second lateral channel (R2) a busy signal, while the transmission of data is prohibited for any station whose radio transceiver device receives such a busy signal on its second side channel. 21. Method according to one of claims 17 to 20, characterized in that, in the absence of request for local transmission and reception, any station transmits artificial messages (ARCC; ADF), in order to maintain a mutual calibration of the pilot frequencies of the different stations. 22. The method of claim 21, characterized in that at least some of the artificial messages are used for service purposes. 23. Method according to one of claims 17 to 22, characterized in that each station can transmit its own messages or relay those of others, according to information called routing tables, defining these transmission relays. 24. The method as claimed in claim 23, characterized in that each station maintains a list of the stations which it receives, and transmits it as a message, which allows it conversely to maintain a list of the stations which receive it. 25. The method of claim 24, characterized in that each station also maintains  - a topological directory designating at least part of the stations with which it can communicate, directly or indirectly, and defining the route or routes of this communication, and  - a routing table defining at any time a preferential route.