WO2006011866A2 - Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau - Google Patents

Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau Download PDF

Info

Publication number
WO2006011866A2
WO2006011866A2 PCT/US2004/020400 US2004020400W WO2006011866A2 WO 2006011866 A2 WO2006011866 A2 WO 2006011866A2 US 2004020400 W US2004020400 W US 2004020400W WO 2006011866 A2 WO2006011866 A2 WO 2006011866A2
Authority
WO
WIPO (PCT)
Prior art keywords
radio
cognitive
channel
wsga
csm
Prior art date
Application number
PCT/US2004/020400
Other languages
English (en)
Other versions
WO2006011866A3 (fr
Inventor
Christian J. Rieser
Thomas W. Rondeau
Charles Bostian
Walling R. Cyre
Timothy M. Gallagher
Original Assignee
Virginia Tech Intellectual Properties, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Virginia Tech Intellectual Properties, Inc. filed Critical Virginia Tech Intellectual Properties, Inc.
Priority to PCT/US2004/020400 priority Critical patent/WO2006011866A2/fr
Publication of WO2006011866A2 publication Critical patent/WO2006011866A2/fr
Publication of WO2006011866A3 publication Critical patent/WO2006011866A3/fr

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
    • H04L1/0002—Systems modifying transmission characteristics according to link quality, e.g. power backoff by adapting the transmission rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/0003—Software-defined radio [SDR] systems, i.e. systems wherein components typically implemented in hardware, e.g. filters or modulators/demodulators, are implented using software, e.g. by involving an AD or DA conversion stage such that at least part of the signal processing is performed in the digital domain
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/38—Transceivers, i.e. devices in which transmitter and receiver form a structural unit and in which at least one part is used for functions of transmitting and receiving
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
    • H04L1/0009—Systems modifying transmission characteristics according to link quality, e.g. power backoff by adapting the channel coding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
    • H04L1/0015—Systems modifying transmission characteristics according to link quality, e.g. power backoff characterised by the adaptation strategy
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/20—Arrangements for detecting or preventing errors in the information received using signal quality detector
    • H04L1/203—Details of error rate determination, e.g. BER, FER or WER
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W8/00—Network data management
    • H04W8/22—Processing or transfer of terminal data, e.g. status or physical capabilities
    • H04W8/24—Transfer of terminal data
    • H04W8/245—Transfer of terminal data from a network towards a terminal

Definitions

  • the present invention generally relates to radio systems providing wireless voice and data links in networks, and more particularly, to a cognitive radio engine architecture which is capable of working under changing and unanticipated circumstances and in the presence of hostile jammers and interferers.
  • the cognitive radio engine is capable of continuously adapting to its environment to conserve resources, such as radio frequency spectrum and battery power, in those applications where those resources are at a premium.
  • a cognitive radio can sense its environment and alter its technical characteristics and operational behavior to benefit both itself and its geographical and spectral neighbors. The ability to sense and respond intelligently distinguishes cognitive radios from fixed or adaptive radios.
  • An adaptive radio can responds to channel conditions that represent one of a limited set of anticipated events.
  • Adaptive radios use artificial intelligence (AI) algorithms that are basically a series of "IF, THEN, ELSE” algorithms. The radio may fail to take any useful action if it meets conditions that its designers never anticipated.
  • AI artificial intelligence
  • a cognitive radio can respond intelligently to an unanticipated event; i.e., a wireless environment (channel) that it never encountered before.
  • the result is enhanced performance (throughput, quality of service (QOS), and security) for the cognitive radio's network and reduced interference to other networks.
  • QOS quality of service
  • a cognitive radio architecture which is based on biologically based models of cognition inspired by child development theories of two-way associative learning through play.
  • the cognitive model of the invention imitates the ability of young minds to adapt rapidly to new situations. Genetic algorithms are well suited for this task because of their ability to find global solutions to changing solution spaces that are often quite irregular. Genetic algorithms are (a) able to synthesize best practices through the crossover operation and (b) enable spontaneous inspiration and creativity through the mutation operation.
  • the multi-tiered genetic algorithm architecture of the invention allows sensing of a wireless channel at the waveform or symbol level, on- the-fly evolution of the radio's operational parameters, and cognitive functions through use of a learning classifier, meta-genetic algorithm, short and long term memory and control.
  • the invention has application in military and disaster communications, where radio systems must work under changing and unanticipated circumstances and in the presence of hostile jammers and interferences.
  • the invention also has application in civilian radio communication systems such as cellular telephones, where spectrum and battery power are at a premium and in which the radios must continuously adapt to conserve these resources.
  • Figure 1 is a block diagram of an implementation of the cognitive radio engine in a network
  • Figure 2 is a concept-level block diagram of the cognitive radio engine of the invention
  • FIG. 3 is a system-level block diagram of the cognitive radio engine of the invention.
  • FIG. 4 is a system-level flowchart of the operation of the cognitive radio engine of the invention.
  • FIG. 5 is a block diagram of the cognitive system monitor (CSM) according to the invention
  • Figure 6 is a detailed block diagram of the CSM according to the invention
  • FIG. 7 is a flowchart of the operation of the CSM according to the invention.
  • FIG 8 is a flowchart showing the "Learning Channel Classifier" routine in the process of Figure 7;
  • FIG. 9 is a flowchart showing the "Update LTM" routine in the process of Figure 7;
  • Figure 10 is a flowchart showing the "Populate STM" routine in the process of Figure 7;
  • Figure 11 is a flowchart showing the "Radio Performance Read” routine in the process of Figure 7;
  • Figure 12 is a flowchart showing the "Goal Optimizer" routine in the process of Figure 7;
  • Figure 13 is a flowchart showing the "Transmit Goal" routine in the process of Figure 7;
  • Figure 14 is a block diagram of the wireless channel genetic algorithm (WCGA) according to the invention;
  • Figure 15 is a block diagram illustrating a specific instantiation of the WCGA for modeling bust errors with genetic algorithm (GA) trained ⁇ Hidden Markov Models (HMMs);
  • Figure 16 is a flowchart of the operation of the WCGA according to the invention.
  • Figure 17 is a flowchart showing the operation of the "Calculate Statistics” routine in the process of Figure 7;
  • Figure 18 is a flowchart of the "Initialize Population” routine in the process of Figure 17;
  • Figure 19 is a flowchart of the "Evaluate Member" routine in the process of Figure 17;
  • Figure 20 is a flowchart of the "Deconstruct Chromosome" routine in the process of Figure 17;
  • Figure 22 is a block diagram of the WSGA according to the invention.
  • Figure 23 is a flowchart of the operation of the WSGA according to the invention.
  • Figure 24 is a flowchart of the "Initialize Population" routine in the process of Figure 23;
  • Figure 25 is a flowchart of the "Evaluate Member" routine in the process of Figure 23;
  • Figure 26 is a flowchart of the "Replace Population" routine in the process of Figure 23;
  • Figure 27 is a block diagram of a distributed cognitive engine according to the invention.
  • Figure 28 is a data map of a standard TCP/IP packet;
  • Figure 29 is a data flow diagram showing the event sequence for transmission of a packet from the WCGA to the CSM;
  • Figure 30 is a datagram of the WCGA information packet sent to the CSM in a TCP payload;
  • Figure 31 is a data flow diagram showing the event sequence for transmission of a packet from the CSM to the WSGA;
  • Figure 32 is a datagram of the WSGA information packet sent to the CSM in a TCP payload;
  • Figure 33 is a data flow diagram showing the event sequence for the transmission of a packet from the WSGA to the CSM.
  • Figure 34 is a datagram of the CSM information packet sent to the CSM in a TCP payload.
  • the network comprises a first plurality of broadband users 1O 1 to 1O n connected to a first router 11 and a second plurality of broadband users 12, to 12 n connected to a second router 13.
  • the first and second routers 11 and 13 are connected by means of a cognitive radio link 15 established by a first adaptive radio 16 controlled by a cognitive radio engine 16a connected to router 11 and a second adaptive radio 17 controlled by a cognitive radio engine 17a connected to router 13.
  • the router 11 is connected to a general internet protocol (IP) network, represented by a cloud 18.
  • IP internet protocol
  • FIG. 19 shows in block diagram form the conceptual components of the cognitive engine. These include a monitor and control component 21, the function of which is to exploit trends, an adaptation and process component 22, the purpose of which is learning, and resources components 23 and 24, respectively concerned with the environment and the radio.
  • Figure 3 shows in more detail the architecture of the cognitive engine.
  • the channel 30, providing data and RF signals, is connected to the radio hardware 32 which, in turn, is connected to the cognitive engine 34.
  • the radio hardware 32 in addition to the usual RF and conversion circuitry, includes a channel estimation process 321 and a radio baseband processor 322.
  • the output of the channel estimation process 321 is input to the wireless channel genetic algorithm (WCGA) module 341 of the cognitive engine 34.
  • This module generates an HMM channel model which is input to the cognitive system monitor (CSM) 342.
  • the CSM 342 also receives radio parameters and performance statistics from the radio baseband processor 322.
  • the CSM 342 generates evaluation functions, associated weights, child chromosomes, and templates which are input to the wireless system genetic algorithm (WSGA) module 343.
  • the evaluation functions, weights, child chromosomes, and a template are referred to as the WSGA information packet.
  • the WSGA 343 generates radio parameters which are applied to the radio baseband processor to adapt the radio to the changing environment.
  • Figure 4 shows the process implemented by the cognitive engine.
  • the process begins by initializing the radio with default settings in the initialization block 401.
  • the WCGA 341 collects information from the channel estimation process 321 and models the channel in function block
  • a test is made in decision block 403 to determine if the channel model is complete. If not, the process loops back to function block 402, but if the model is complete, the channel model is passed to the CSM 342 in function block 404. In function block 405, the CSM 342 determines if a new radio configuration is needed, builds trends from the WCGA 341 and radio statistics, and develops a WSGA information packet for the WSGA 343. A test is made in decision block 406 to determine if the system needs a new configuration. If not, the process loops back to function block 405; otherwise, a WSGA information packet is passed to the WSGA 343 in function block 407.
  • the WSGA 343 develops a system chromosome and creates fitness from mathematical idealization of radio performance.
  • a test is then made in decision block 409 to determine if the new system developed by the WSGA 343 is close to optimization. If not, the process loops back to function block 408; otherwise, the new system configuration is passed to the baseband radio system 322 in function block 410.
  • the baseband radio system 322 then reconfigures itself in function block 411. Thereafter, the radio system monitors its performance and collects statistics (bit error rate (BER), data rate, etc.) in function block 412. These radio performance parameters are collected in function block 413 and input to the CSM at function block 405.
  • BER bit error rate
  • data rate data rate
  • FIG. 5 is a block diagram of the CSM showing the functional relationships of the several components of Figure 3.
  • the radio hardware here shown as block 50, generates radio feedback parameters 501 which are input to the CSM, here shown as block 51.
  • the CSM 51 is also connected to the WCGA and WSGA, here shown as blocks 52 and 53, respectively. More particularly, the WCGA 52, through its TCP/IP
  • the CSM 51 includes, among other things, a knowledge base (KB) in long term memory (LTM) 513.
  • the CSM 51 classifies observed environment (i.e., the channel model), updates the knowledge base in long term memory, and synthesizes an information packet for the WSGA (e.g., fitness function, initial population, weights, and templates).
  • TCP_KB_PORT is an optional feature to allow socket- layer communications with LTM stored on a remote radio system in a distributed implemenation.
  • FIG. 6 shows in more detail the structure of the CSM 51, and in this figure like reference numerals to those in Figure 5 represent the same of similar components.
  • An observed channel and location buffer 601 receives input from the WCGA 52 via the TCP/IP socket layer 511 and radio parameters from the radio and WSGA performance API (application program interface) 602. The data in this buffer is supplied to a channel statistics processor 603. The statistics computed by the processor 603 are input to a learning channel classifier 604 which, in turn, provides outputs to a short term memory (STM) 605 and a channel updater 606. The channel updater 606 generates a pointer to a TCP/IP link converter in the cognitive knowledge base in the long term memory (LTM) 513.
  • the LTM long term memory
  • the goal evolver 607 also receives input from the radio and WSGA performance API 602 and provides an output to the estimated radio goal and location buffer 608. The data in this buffer is provided to the WSGA 53.
  • the CSM 51 is an evolutionary algorithm that is the core of the cognitive radio intelligence and directs all knowledge gained from the sensing processes to the adapting processes.
  • the CSM 51 is a meta-GA (genetic algorithm), serving as short and long term memory to understand and utilize past trends in system behavior.
  • the CSM 51 is the creative side of the cognitive radio brain.
  • the WSGA 53 takes direction from the CSM 51 and performs its own genetic algorithm (GA) to redesign the radio configuration.
  • the WSGA 53 is the logical side of the cognitive radio brain.
  • the WSGA 53 acts upon a chromosome containing parameters like transmitter power, frequency, bandwidth, modulation, channel coding, etc.
  • the WSGA 53 also receives the fitness evaluation to direct the algorithm in how it should attempt to evolve.
  • the fitness could be a measure of the minimum BER, minimum power consumption, maximum carrier to noise ratio (C/N), minimum latency, maximum data rate, or a combination of these.
  • the GA will attempt to optimize these fitness functions.
  • the CSM 51 receives from the system and the WCGA 52 sensing algorithm, the fitness function may change to achieve new goals.
  • the WSGA 53 works within the knowledge established by the CSM 51 to find a configuration that should work.
  • the WSGA 53 then passes the new radio configuration to the RF and digital baseband equipment, which then reconfigures itself. After the radio tries the new configuration, the CSM 51 will evaluate the radio's performance to assess the success of the configuration. If the CSM 51 determines that the radio's performance is not optimum, it will inform the WSGA 53 that a new configuration is needed.
  • the task of the WCGA 52 is to sense the environment and model it as accurately as possible to provide information about the multipath environment or other interferences that can cause errors in the channel.
  • the environmental information is then sent to the CSM 51 for processing.
  • the CSM 51 can then use the model to evaluate what functions are applicable to the current channel. Based on the environment, the CSM 51 can determine how best to analyze the system configuration, what performance to attempt to achieve, and how best to set the fitness evaluation function for the WSGA 53 so that the fitness of the system chromosome will converge upon the appropriate radio settings.
  • the WCGA 52 first models the channel and then the CSM 51 estimates the maximum data rate achievable within the system given its knowledge of that channel.
  • the CSM 51 will inform the WSGA 53 to use the data rate as its fitness evaluation and pass the maximum data rate as the fitness goal to achieve.
  • the CSM 51 will also tell the WSGA 53 that it is limited by the frequency and bandwidth of the UNII band.
  • the WSGA 53 will perform the GA to produce new system configurations and test them based on two metrics: does the new configuration meet the regulatory requirements and does it achieve the data rate requirement?
  • the WSGA 53 When the WSGA 53 is satisfied that the regulatory requirements are met and that the data rate requirement is close to being achieved based on the fitness evaluation that it was provided by the CSM 51, it sends the information to the radio for reconfiguration. Artificial intelligence researchers would consider the WSGA 53 as working with "simulations", while the CSM 51 monitors "reality”, which means that the simulated results will often deviate from the actual radio performance. Therefore, the CSM 51 will have to monitor how the radio performs so that it can instruct the WSGA 53 if a new configuration is required.
  • the process implemented by the CSM 51 is illustrated in the flowchart of Figure 7, to which reference is now made.
  • the process begins in input block 701 where the observed channel from the WCGA 52 is read into the observed channel and location buffer 601.
  • the channel metric calculator 603 then calculates the ranking metric of the observed channel in function block 702.
  • the ranking metric can be anything that distinguishes the channel like average bit error rate or average burst error length, which is the current metric being used.
  • the learning channel classifier 604 finds the closest match to the observed channel in long term memory (LTM) in function block 703. This is done by GA channel index search by the ranking metric or by binary tree search by the ranking metric. Any change is indicated to the goal evolver 607 via the short term memory (STM) 605.
  • the LTM 513 is updated in function block 704.
  • the STM 605 is populated in function block 705 with knowledge base chromosomes in
  • radio performance parameter and existing WSGA simulation fitness, population, tags and templates are read into the goal evolver 607.
  • Goals in the STM 605 are crossed over or mutated with estimated radio goal for the observed channel in function block 707.
  • decision block 708 a test is made to determine if the optimal goal has been chosen. If not, the process loops back to function block 707, but if so, the data in estimated radio goal and location buffer 608 is formatted into an information packet for the WSGA 53 in function block 709. The formatted packet is then transmitted to the WSGA 53 in output block 710, and the process returns to input block 701.
  • the algorithm is a never-ending loop while the radio is operational to provide continuous evolution of the radio's settings to either continually search for a more optimal solution, or to adapt the radio in a time-varying channel.
  • Figure 8 is the flowchart for the learning channel classifier routine of function block 703 in Figure 7.
  • the input 801 is the computed ranking metric from the channel statistics processor 603.
  • the channel statistics are converted to channel cluster index 7, using best fit weights, in function block 802.
  • the statistics of the observed channel are compared with those of the corresponding channel stored in long term memory 513 in function block 803.
  • a test is made in decision block 804 to determine if the comparison matches within a predetermined difference (delta). If so, the location of the estimated channel j is stored in register m, a pointer in the CSM to a member of LTM 513, in function block 805; otherwise, a new candidate is selected from LTM 513 in function block 806, and a return is made to function block 803.
  • Figure 9 is the flowchart for the update LTM routine in function block 704 in Figure 7.
  • the input 901 is new member.
  • a ranking metric of the observed channel is calculated in function block 902.
  • a test is made in decision block 903 to determine if the current LTMLength is equal to zero. If it is, the observed channel is placed in the first position of LTM in block 904 and the function exits at 905. Otherwise, the length of LTM is compared the MaxLTMLength, the maximum possible length of the LTM member, in decision block 906. If the current length is less than the maximum length, function block 907 determines what member of the population, pointed to by register m has the closest matching ranking metric to the observed channel.
  • the function exits 909. If the metrics are the same as determined in decision block 908, there is no need to update LTM, so the function exits 909. If the metrics are different, the observed channel is inserted into LTM at position pointed to by m at block 910. The current LTM length is then incremented by function block 911. It is important to note that the channel that matched closest to the observed channel is not replaced, the new channel is just inserted at this position to maintain proper order based on the ranking metric in LTM.
  • block 913 finds the closest matching member of LTM like block 907 did.
  • the differences in the metrics between the observed channel and the two surrounding channels of LTM member m is calculated in block 914 as ⁇ 0 and the difference between LTM member m and its two surrounding neighbors is calculated as ⁇ m , where ⁇ is the Euclidean distance between the LTM members surrounding m.
  • Decision block 915 determines if ⁇ 0 is less than ⁇ m , whereby the current member of LTM at position m is replaced by the new channel in block 916 and the function exits at 917; otherwise, the current LTM member is not replaced and the function exits at 917. This algorithm ensures that LTM exists of observed channels that maximize the distance between any two members to provide as wide a scope of the possible channels as possible.
  • Figure 10 is the flowchart for the populate STM routine in function block 705 in Figure 7.
  • the register m as developed in block 704 is passed to this function in block 1001. If the current LTM length is less than MaxSTMLength, the maximum number of members short term memory (STM) can hold, as decided in block 1002, then block 1003 sets CurrentSTMLength, the current length of STM, to the current length of LTM. Otherwise, block 1004 sets the current STM length to the maximum
  • Block 1005 begins a loop which takes LTM members at m minus i, an index value that starts at 0 and increments in steps of 1 to CurrentSTMLength/2, and places them into STM in block 1006.
  • Block 1007 initializes j, another index value, to zero. Index i is then set in block 1008 to the middle of the current STM member to populate the upper half of STM.
  • FIG. 11 is the flowchart for the read radio performance parameters routine in function block 706 in Figure 7.
  • the inputs 1101 are the radio feedback parameters read by the radio performance API (application program interface). These include such things as BER (bit error rate) power, battery life, etc.
  • the radio performance API stores the radio parameters in a buffer to be read by the goal evolver 607 in function block 1102.
  • the output 1103 is a signal to the goal evolver 607 that new radio performance values are available.
  • Figure 12 is the flowchart for the goal optimizer routine in function block 707 and decision block 708 in Figure 7.
  • the inputs 1201 are the radio performance parameters and existing WSGA simulation fitness, population, and templates.
  • Block 1202 compares the simulated meters, / WSGA w * m the actual observed parameters from the radio, / radl0 . The differences shows how far the WSGA simulation was from the actual operation of the radio. From here, it can be determined how necessary a new configuration is required, or learn about behavior to avoid or promote in the future.
  • Block 1203 ranks the members of STM in respect to two objectives: similarity and utility (drawing from decision theory vocabulary).
  • the similarity functions determine how close the observed channel is to each member in the short term memory, and the utility function determines how successful each member of STM has been in the past. Poor performing members of memory lose worth over time and successful members gain worth.
  • the candidate member of STM is that member who maximizes both similarity and utility.
  • the goal vectors refer to the information sent to the WSGA about what the channel is and how it should be evaluated (fitness functions, weights, template, and child chromosomes).
  • the STM then undergoes manipulation through a genetic algorithm in block 1204, which is an evolutionary process that attempts to alter the goal vectors of the poor performing channels by combining successful pieces of goal vectors from other members of STM.
  • Block 1205 illustrates how a member of STM may look, with the channel on the left, which is used to determine the similarity function, and goals in the vector on the right.
  • the genetic algorithm in 1204 attempts to alter the goal vectors of the listed channel by combining goals A, B, C, and D from different members of STM. These become the new goal vector associated with that channel, which will hopefully produce a better utility function than the previous goal vector. This process attempts to "play" with past behavior to learn better ways of behaving in the future.
  • Block 1207 shows the WSGA sending the new system chromosome back to the CSM as well as the simulated parameters.
  • Block 1208 shows the radio transmitting the actual parameters associated with the new system chromosome back to the CSM.
  • Blocks 1207 and 1208 run independent of the CSM, and so they must be asynchronously timed where the information coming from either entity (the WSGA or the radio) can come at any given time.
  • the radio parameters are returned, they are compared to the simulated parameters by observing the differences between the simulated and the actual parameters at function block 1209. This information is then used by block 1210 to update the worth associated with the STM member used to generate the current system chromosome. If the real system performs worse than the simulation, it indicates a problem with the STM goal vector, and the worth of that STM member is decreased.
  • FIG. 13 is the flowchart for the transmit goal routine in output block 710 in Figure 7.
  • the input 1301 is goal in buffer 608.
  • a TCP (transport control protocol) socket packet with WSGA information is built in function block 1302.
  • a socket is created in function block 1303, and the created socket is bound to an unused local port in function block 1304.
  • the socket is then connected to a remote port in function block 1305, and the TCP packet is sent with the WSGA information to the remote system in function block 1306.
  • the process ends by closing the connection in function block 1307 and exiting at block 1308.
  • the wireless channel genetic algorithm (WCGA) is, in general, a mechanism by which the channel is observed and modeled for use in the CSM.
  • FIG 14 is a block diagram of the WCGA.
  • a channel capturing device 1401 i.e., the channel estimation process 321 in Figure 3
  • the channel capturing device can be any known method of modeling a channel using devices such as a channel sounder, a frequency domain capturing device (Fast Fourier Transform device), training data, etc.
  • the WCGA creates a machine-usable model of the radio channel. This is transmitted via the TCP/IP socket layer 521 to the corresponding TCP/IP socket layer 511 of the CSM 51.
  • FIG. 15 A specific instantiation of the WCGA is shown in Figure 15. This instantiation models burst errors with GA-trained HMMs.
  • the WCGA uses an error stream for the input, which is a train of symbols representing the number of bit errors per symbol. For the WCGA to produce an accurate model, many thousands of error symbols must be collected, which would require a long training sequence, taking both time and bandwidth.
  • a more compact and efficient approach to channel modeling is to utilize the information collected by the channel sounder. While the channel sounder response can provide an immediate understanding of the channel, the data received from the sounder is large and bulky. By using the channel sounder response, a model of the channel is derivable by simulating the channel as a filter with an impulse response derived from the channel sounder.
  • a random bit sequence passed through the simulated channel will produce an error sequence. Because we are interested in a statistical model of the channel, we can use the simulated channel instead of the error sequence.
  • the Hidden Markov Model (HMM) of the channel developed by either a true error sequence or a simulated error sequence is still a statistical representation of the channel. However, this representation is very small compared to the channel sounder data and is capable of representing the channel equally well.
  • the channel, 1501 is received by a channel sounder 1502, in this case the broadband channel sounder developed at the Center for Wireless Telecommunications. This channel sounder uses impulse transmitted from one radio and is captured by the sounder on another radio using a sliding correlator sounding technique, which captures a small amount of energy from the pulse at different offsets.
  • the channel sounder response is in the form of a waveform representative of the transmitted impulse, and which contains information about the channel.
  • Block 1503 takes this impulse and converts it to a linear time invariant (LTI) model of the channel, h(t), by down-converting the signal to baseband. Simulated data is then convolved through the LTI model in block 1504, out of which an error stream comes.
  • the error stream represents what transmitted symbols would be received as good symbols and which would be error symbols.
  • Block 1505 takes the error stream and calculates statistics stored in Observed_Histogram of block 1506.
  • Block 1507 then generates a randomly initialized WCGA chromosome population according to Figure 18, which is used to start the genetic algorithm (GA) of block 1508.
  • the GA performs standard crossover and mutation operations and evaluates each chromosome against the observed channel histogram in block 1508 by creating a channel histogram from the HMM and finding the sum of the differences between each point in the histogram as described in Figure 19. The best fitness is the one closest to zero.
  • the GA exits and transmits the final HMM model to the CSM via the TCP/IP socket layer of block 521.
  • Figure 16 is a flow chart of the process implemented by the WCGA.
  • the input 1601 is the channel information.
  • the CWT's Broadband sounder in block 1602, as described by block 1502, is used to create a channel impulse response, from the sampled pulse.
  • block 1603 develops the mathematical representation, h(t), of the channel and block 1604 uses h(t) to generate an error stream.
  • Block 1605 like 1505, calculates the channel statistics as described in Figure 17, and the histogram from the channel statistics is stored in block 1606 as the Observed_Histogram.
  • Block 1607 randomly initializes a population of
  • HMM chromosomes according to the routine of Figure 18.
  • Block 1608 starts the genetic algorithm loop, which runs until a specified stopping criteria like a limited number of generations (the currently used method), or a certain minimum desired fitness value.
  • a selection process is used in block 1609, which can be any GA selection process such as tournament or roulette wheel, to choose parents for mating.
  • the parents are then genetically manipulated through crossover, block 1610, and mutation, block 1611, to create a new set of offspring.
  • the offspring are then evaluated in block 1612 according to the routine in Figure 19.
  • the worst members of the current generation are then replaced by more fit offspring in block 1613, and the entire population is evaluated in block 1614 based on the fitness values developed in block 1612.
  • the best fit member of the population can then be used in 1608 to determine if the stopping criteria is met.
  • FIG. 17 is the flowchart for the calculate channel statistics routine of function block 1605 in Figure 16.
  • the input 1701 is the error steam from the observed channel and location buffer 1701.
  • the number of symbols in the stream is counted 1703, and loop 1704 is used to count the number of bits represented in the error stream
  • the loop initiated in 1708 tests if any bit in each symbol is an error through decision block 1709; all errors are added to the total number of bit errors 1710. If the symbol does not contain any errors as determined in 1706, the current symbol is compared to the previous symbol 1711. If they are different, the burst histogram for the specified burst length is incremented by block 1712. Block 1713 decides if the current burst length is the longest burst length observed; if so, the longest burst length observed, as stored in
  • MaxBurstLength is updated with the new maximum burst length in block 1714.
  • Block 1715 resets the burst length value to 1 for the next burst.
  • This algorithm is used to develop statistics of the channel including the maximum burst length, the number of symbols and bits, the number of symbol and bit errors, the symbol error rate, the bit error rate, and the histogram of bursts in the channel as stored in block 1716.
  • Figure 18 is a flowchart of the initialize population routine, 1801, in function block 1603 in Figure 16.
  • a loop of i 0 to length N, the number of states in the HMM, is run from block 1802.
  • Each index; of row i is set to a random floating point number from 0 to 1 in block 1805 and block 1806 increments the counter by the random value generated in block 1805.
  • the entire row i in the A matrix has been filled with values. Since this is a probability matrix, all of the elements in each row must sum to 1; therefore, block 1807 divides all elements of row i by the sum of all the elements in that row.
  • the final row vector now sums to 1.
  • Loop 1809 this time runs fromy to M, the number of outputs possible at all states (i.e., good symbols and bad symbols for an M of 2).
  • Block 1810 sets the elements of row i of matrix B, 2102, to a randomly generated number from 0 to 1, and, again, the counter is incremented in block 1811.
  • the vector rows of matrix B are then normalized in block 1816 to 1 by dividing all the elements by the sum of all the row elements.
  • Loop 1802 ends, and the Pi initialization vector of the HMM,
  • the counter is again set to 0 at block 1812, and loop 1813 cycles for j to N to fill all elements of the N-length Pi vector.
  • Block 1817 normalizes the vector so that the elements sum to 1.
  • FIG. 1818 simply converts all matrices of the HMM, A, B, and Pi, into a single vector to represent the chromosome.
  • Figure 21 shows the HMM in terms of matrices, 2101, 2102, and 2103, and as a chromosome, 2104.
  • Figure 19 is a flowchart of the evaluate member routine in function block 1608 in Figure 16.
  • the evaluate routine, 1901 first deconstructs the chromosome back into its HMM form in block 1902 so the HMM can be used to generate the error stream properly in block 1903.
  • the channel statistics are then calculated in block 1904 according to Figure 18.
  • Figure 20 is a flowchart of the deconstruct chromosome routine of 1902.
  • the routine takes in a chromosome at block 2001 and sets an index value to 0 in block 2002.
  • Each element in matrix A is a flowchart of the deconstruct chromosome routine of 1902.
  • the routine takes in a chromosome at block 2001 and sets an index value to 0 in block 2002.
  • Each element in matrix A
  • A[i][/ ' ] is placed into the chromosome vector at position index, which is then incremented, in block 2005.
  • Matrix A, 2101 is the NxN state transition matrix of the HMM. Give a current state, there is a certain probability of moving to any other of the states or of staying in the same state. The columns set the current state and the row sets the state being transitioned to. For example, element A21 is the probability of going from state 1 to state 2.
  • Matrix B, 2102 is the NxM state output matrix of the HMM. At any given state, represented by the row, there is a probability of outputting a certain output symbol, represented by the column.
  • the output values can represent a good or bad bit, or a good or bad symbol (where a symbol can represent many bits). For example, an output of zero represents a good symbol and an output of one represents a bad symbol, so given that the HMM is in state 1 at a given time, there is a probability of BIl of outputting a 0, or good symbol, and a probability of B 12 of outputting a 1, or bad symbol.
  • the chromosome representation in 2104 shows how the HMM is converted into a chromosome for manipulation by the genetic algorithm. Each row and matrix is lined up back to back to create a single vector that can then be used jn genetic operations such as crossover and mutation.
  • Figure 22 is a detailed block diagram of the WSGA 53. It receives input from the CSM 51 via the TCP/IP socket layer 531. Information about the WSGA is stored in a structure called WSGAInfo, and the member chromosomes of the genetic algorithm are stored in 2201. Block 2202 initializes the member chromosomes as shown in Figure 24.
  • Block 2203 is the genetic algorithm used to determine the new radio system parameters as shown in Figure 23, which links to a dynamic link library (DLL) to retrieve the mathematical fitness functions in block 2204.
  • DLL dynamic link library
  • the final solution from the genetic algorithm is transmitted to the radio via a radio-specific Application Programmable Interface (API) of block 2205.
  • API Application Programmable Interface
  • the WSGA chromosome structure is shown in the following table.
  • This table is viewed as a vector in the algorithm and operated upon as a chromosome through genetic algorithm crossover and mutation procedures.
  • the values of these table parameters determine the fitness of the chromosome and the behavior of the radio.
  • Figure 23 is a flowchart showing the operation of the WSGA.
  • the input 2301 is a packet from the CSM 51 which is temporarily stored at 2302.
  • the population of chromosomes is initialized in 2303 according to the routine of Figure 24.
  • Decision block 2304 controls the genetic algorithm loop and exits the loop upon a stopping criteria, which could be a certain number of generations or after a decrease in performance gain per generation is detected (that is, the fitness of the current generation did not differ significantly from the previous generation).
  • block 2305 selects parent chromosomes that will be used to generate offspring chromosomes to replace the population the next generation.
  • Blocks 2306 and 2307 perform standard genetic algorithm techniques of crossover and mutation, respectively.
  • Block 2308 evaluates the fitness values for each chromosome, both parent and offspring.
  • Block 2309 determines which members of the population to replace using a relative fitness evaluation method of Figure 26. Once the genetic algorithm loop has exited, block 2310 transmits the system parameters as defined in the best fit chromosome of the final generation to the radio via an API.
  • Block 2311 also transmits the best fit chromosome along with the simulated fitness values to the CSM so the CSM can compare the simulated fitness values to the real fitness values read from the radio after the new radio settings have been set.
  • Figure 24 is the flowchart of the initialize population routine of function block 2303 in Figure 23. This routine first fills the initial population of the WSGA with children received from the CSM and then randomly generates any more children required to fill the population. Block 2401 initializes the routine. Loop 2402 cycles through all child chromosomes received from the CSM, where block 2403 inputs the child into the population and block 2404 evaluates the chromosomes fitness values. Loop 2405 then cycles through the remaining populated indices where block 2406 randomly generates chromosomes to fill the population and block 2407 evaluates the fitness of the new chromosomes. Once the population is filled, the routine exits 2408.
  • FIG. 25 is a flowchart of the evaluate member routine of function block 2308.
  • the chromosome is translated to absolute radio parameters (power in terms of dBm, frequency in Hz, etc.) stored in structure data through the radio-specific API at block 2502.
  • Loop 2503 cycles through all of the current fitness functions used to evaluate the members using an index variable i.
  • each fitness function is evaluated by calling the function out of the "WSGAFitFunc.dll" dynamic link library (DLL) in block 2504.
  • the DLL is useful for dynamic linking because the radio system may be updated in real-time with new or improved fitness evaluation functions without altering the rest of the system.
  • Block 2505 uses the function from the DLL to calculate the fitness for each fitness function, called an objective by passing the data structure to the function as well as a meters structure, which is updated inside the DLL function to contain simulated meters of the radio's performance.
  • the routine exits 2506.
  • Figure 26 is a flowchart of the replace population member routine
  • Loop 2601 based on a relative tournament selection scheme.
  • Loop 2602 cycles through by incrementing index i from 0 to the total population size.
  • Each cycle 2 members, designated as member[n] and member[&], are chosen from the population of parents and offspring at block 2603.
  • Loop 2604 then uses index j to cycle through all of the fitness functions.
  • Decision blocks 2605, 2606 and 2607 decide which of the two members won by comparing the fitness values associated with each objective. If the objective of member[n] is greater than the objective of member[&] in block 2605, then member[rc] wins and block 2608 increments the fitness value by adding the amount of weight associated with the current fitness function being compared to the member's fitness.
  • members [n] and [k]'s fitness is compared in block 2611, and if member[n]'s fitness is larger than member[£]'s fitness, then member[&] is removed from the population and member[n] survives in the population to be a part of the next generation in block 2612; otherwise, member[n] is killed and member[fc] survives to the next generation in block 2613.
  • This relative tournament selection mechanism is a way to compare population members and choose the best fit members for survival to the next generation when there are multiple objectives to consider.
  • the weights allow the system to adjust its priorities when it thinks one fitness function is more important than another (e.g., minimizing the bit error rate may be more important than maximizing the data rate, so the weights can help determine how much each matters).
  • the CSM evolutionary algorithm consists of a learning classifier function that classifies the observed channel model received from the WCGA or broadband channel sounder and a meta-genetic algorithm that determines the appropriate fitness function, chromosome structure, and templates using the crossover operator based on knowledge from its short and long term memories as well as the creative new solutions generated from its mutation functions.
  • the genetic algorithm (GA) approach to adapting a wireless radio provides many benefits.
  • GA genetic algorithm
  • it is a chaotic search with controllable boundaries that allow it to seek out and discover unique solutions efficiently.
  • chaotic behavior could produce a solution that is absolutely correct but may be counter-intuitive.
  • the cognitive system can ensure legal and regulatory compliance as well as efficient searches.
  • the cognitive system defines the radio chromosome, where each gene represents a radio parameter such as transmit power, frequency, modulation, etc.
  • the adaptation process of the WSGA is performed on the chromosomes to develop new values for each gene, which is then used to adapt the radio settings. If a radio cannot adjust a particular parameter, then the adaptation process will ignore the gene representing the parameter. Also, if there are certain parameters unique to a particular adaptable radio, a few genes can be left unused so as to be used for such proprietary purposes. See, again the Chromosome Parameters table above.
  • each radio will have a unique method of adapting the radio parameters and each parameter will mean something different, a small hardware interface is required to connect the WSGA to the radio.
  • the hardware interface will take the chromosome from the WSGA and use the gene values to properly update the radio.
  • the hardware interface is a small piece of software required for each radio while the cognitive processing engine remains system-independent. While the independence of the WSGA and the cognitive processor to the radio allows any adaptable radio to become a cognitive radio, it should be clear that the more adaptable a radio is, the more powerful the cognition becomes.
  • Each of the three main algorithms (CSM, WSGA and WCGA) can be co-located or distributed. "Co-located” means that the three algorithms exist in the same radio with shared memory and processing. "Distributed” means that one or more of the algorithms exists on another radio with separate memory and processing. The knowledge base developed in long term memory may also be distributed, allowing for physically distributed cognitive consciousness that appear logically the same to the CSM.
  • a base station unit may have both the WCGA and CSM in its system to model the radio environment and maintain the long term memory that the subscriber units can access for their independent WSGA algorithms running locally.
  • This scenario allows for a common memory bank and associated realization of the radio environment, but each system can adapt independently. If a radio adapts to a set of parameters that are more successful in the radio environment, it can then communicate its successful adaptation back to the BSU for better future adaptation by all radios.
  • Another scenario could be that the radio has its own sensing mechanism, but sends the channel model to another radio for CSM and WSGA processing. The output of the WSGA could then be sent back to the original radio for adaptation.
  • the long term memory could be mapped to a local device or set of devices.
  • This distributed knowledge base concept allows the power of the GA approach to be realized, because as the network becomes more complex, the knowledge base has a mechanism to scale with it.
  • the method of exchanging data between the algorithms must allow for both co-located and distributed systems.
  • the interface between the systems is all packet based and is sent using TCP to a specific socket in the system.
  • the basic TCP datagram is illustrated in Figure 28.
  • the socket-based communications is convenient because it has a common port to connect to but is also related to the IP address of the radio system.
  • the IP address can be simply set to the system's ow!n IP address or the internal loopback address (usually defined as 127.0.0.1).
  • a change in the IP address is all that is required.
  • the WCGA When the WCGA has a channel model for the CSM, it opens a TCP socket connected to port WCGA_CSM_TCP_PORT. A packet of information is then sent from the WCGA to the CSM and the connection closed.
  • the sequence of events for passing the data between the WCGA and the CSM is shown in Figure 29.
  • the WCGA-CSM packet for passing channel statistics and a burst error histogram is shown in Figure 30.
  • the CSM to WSGA communications is very similar to the WCGA to CSM communications; however, now the CSM opens the communications when the packet is ready.
  • the CSM opens a TCP socket in the WSGA on part WSGA_TCP_PORT.
  • the packet is sent, and the connection is closed.
  • the sequence of events for passing the data between the CSM and the WSGA is shown in Figure 31.
  • the CSM-WSGA packet for passing the WSGA control information is shown in Figure 32.
  • the WSGA then passes information back to the CSM regarding the new system chromosome developed. Again, the communications are very similar to the previous method, only the WSGA opens a TCP socket on the CSM on port WSGA_CSM_TCP_Port. The packet is sent, and the connection is closed. The sequence of events for passing the data between the WSGA and the CSM is shown in Figure 33. The WSGA-CSM packet for passing the WSGA final information is shown in Figure 34.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Quality & Reliability (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Theoretical Computer Science (AREA)
  • Genetics & Genomics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Physiology (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

L'invention porte sur une approche d'algorithme génétique (GA) permettant d'adapter une radio sans fil à un environnement changeant. Un moteur radio cognitif met en place trois algorithmes : un algorithme génétique de canal sans fil (WCGA), un moniteur de système cognitif (CSM) et un algorithme génétique de système sans fil (WSGA). Une recherche chaotique comprenant des limites contrôlables permet au moteur radio cognitif de chercher et découvrir de manière efficace des solutions uniques. Du fait de sa capacité à contrôler l'espace de recherche par limitation du nombre de générations, de taux critiques, de taux de mutation, d'évaluations de santé physique, etc., le système cognitif peut assurer une conformité légale et réglementaire ainsi que des recherches efficaces. La versatilité du procédé cognitif peut être appliquée sur n'importe quelle radio adaptative. Le système cognitif définit le chromosome radio, chaque gène représentant un paramètre radio tel qu'une puissance de transmission, une fréquence, une modulation, etc. Le procédé d'adaptation de WSGA est appliqué sur les chromosomes afin de développer de nouvelles valeurs pour chaque gène, et sert ensuite à adapter le réglage radio.
PCT/US2004/020400 2004-06-25 2004-06-25 Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau WO2006011866A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/US2004/020400 WO2006011866A2 (fr) 2004-06-25 2004-06-25 Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2004/020400 WO2006011866A2 (fr) 2004-06-25 2004-06-25 Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau

Publications (2)

Publication Number Publication Date
WO2006011866A2 true WO2006011866A2 (fr) 2006-02-02
WO2006011866A3 WO2006011866A3 (fr) 2006-08-10

Family

ID=35786600

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2004/020400 WO2006011866A2 (fr) 2004-06-25 2004-06-25 Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau

Country Status (1)

Country Link
WO (1) WO2006011866A2 (fr)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006096779A1 (fr) * 2005-03-07 2006-09-14 Symstream Technology Holdings Pty Ltd. Procede et appareil a fonctionnement de type organisme pour systeme radio virtuel a flux de symboles
CN100373981C (zh) * 2006-05-11 2008-03-05 电子科技大学 认知无线电中一种混合网络结构的实现方法
WO2008104935A2 (fr) * 2007-02-26 2008-09-04 Nokia Corporation Appareil, procédé et produit de programme informatique fournissant une sélection de canal de radio cognitive améliorée
WO2008154586A1 (fr) * 2007-06-11 2008-12-18 Broadcom Corporation Procédé et système pour un circuit intégré de communication configurable ou un chipset
US8250007B2 (en) 2009-10-07 2012-08-21 King Fahd University Of Petroleum & Minerals Method of generating precedence-preserving crossover and mutation operations in genetic algorithms
US10405219B2 (en) 2017-11-21 2019-09-03 At&T Intellectual Property I, L.P. Network reconfiguration using genetic algorithm-based predictive models
US10986509B1 (en) 2020-05-14 2021-04-20 At&T Intellectual Property I, L.P. Placement of antennas for fifth generation (5G) or other next generation networks
CN116502976A (zh) * 2023-06-27 2023-07-28 中铁第四勘察设计院集团有限公司 一种多式联运枢纽工艺布局方法及系统

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CHARLES W. ET AL.: 'Rapidly Deployable Broadband Communications for Disaster Response' VIRGINIA TECH. March 2003, pages 1 - 39 *
REISER C.J.: 'Biologically Inspired Cognitive Radio Engine Model Utilizing Distributed Genetic Algorithms for Secure And Robust Wireless Communications and Networking' VIRGINIA TECH., DISSERTATION pages 1 - 168 *
'Special Report: ECEs and Biomedicine' VIRGINIA TECH. April 2004, pages 1 - 4 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8135362B2 (en) 2005-03-07 2012-03-13 Symstream Technology Holdings Pty Ltd Symbol stream virtual radio organism method and apparatus
WO2006096779A1 (fr) * 2005-03-07 2006-09-14 Symstream Technology Holdings Pty Ltd. Procede et appareil a fonctionnement de type organisme pour systeme radio virtuel a flux de symboles
CN100373981C (zh) * 2006-05-11 2008-03-05 电子科技大学 认知无线电中一种混合网络结构的实现方法
WO2008104935A2 (fr) * 2007-02-26 2008-09-04 Nokia Corporation Appareil, procédé et produit de programme informatique fournissant une sélection de canal de radio cognitive améliorée
WO2008104935A3 (fr) * 2007-02-26 2008-10-30 Nokia Corp Appareil, procédé et produit de programme informatique fournissant une sélection de canal de radio cognitive améliorée
US8995288B2 (en) 2007-06-11 2015-03-31 Broadcom Corporation Method and system for a configurable communication integrated circuit and/or chipset
WO2008154586A1 (fr) * 2007-06-11 2008-12-18 Broadcom Corporation Procédé et système pour un circuit intégré de communication configurable ou un chipset
US8250007B2 (en) 2009-10-07 2012-08-21 King Fahd University Of Petroleum & Minerals Method of generating precedence-preserving crossover and mutation operations in genetic algorithms
US10405219B2 (en) 2017-11-21 2019-09-03 At&T Intellectual Property I, L.P. Network reconfiguration using genetic algorithm-based predictive models
US10945145B2 (en) 2017-11-21 2021-03-09 At&T Intellectual Property I, L.P. Network reconfiguration using genetic algorithm-based predictive models
US10986509B1 (en) 2020-05-14 2021-04-20 At&T Intellectual Property I, L.P. Placement of antennas for fifth generation (5G) or other next generation networks
CN116502976A (zh) * 2023-06-27 2023-07-28 中铁第四勘察设计院集团有限公司 一种多式联运枢纽工艺布局方法及系统
CN116502976B (zh) * 2023-06-27 2023-11-03 中铁第四勘察设计院集团有限公司 一种多式联运枢纽工艺布局方法及系统

Also Published As

Publication number Publication date
WO2006011866A3 (fr) 2006-08-10

Similar Documents

Publication Publication Date Title
US7289972B2 (en) Cognitive radio engine based on genetic algorithms in a network
Rieser et al. Cognitive radio testbed: further details and testing of a distributed genetic algorithm based cognitive engine for programmable radios
Rieser Biologically inspired cognitive radio engine model utilizing distributed genetic algorithms for secure and robust wireless communications and networking
O’Shea et al. Approximating the void: Learning stochastic channel models from observation with variational generative adversarial networks
Xu et al. The application of deep reinforcement learning to distributed spectrum access in dynamic heterogeneous environments with partial observations
CN111628855B (zh) 基于深度强化学习的工业5g动态多优先级多接入方法
Bors et al. Minimal topology for a radial basis functions neural network for pattern classification
Baldo et al. Learning and adaptation in cognitive radios using neural networks
CN105790813B (zh) 一种大规模mimo下基于深度学习的码本选择方法
CN107743103A (zh) 基于深度学习的mmtc系统的多节点接入检测和信道估计方法
CN110167176B (zh) 一种基于分布式机器学习的无线网络资源分配方法
CN112910811B (zh) 基于联合学习的噪声水平未知条件下的盲调制识别方法和装置
Berberidis et al. Data sketching for large-scale Kalman filtering
CN110300075A (zh) 一种无线信道估计方法
CN112769510A (zh) 用于基于机器学习来估计干扰的装置及方法
WO2006011866A2 (fr) Moteur radio cognitif fonde sur des algorithmes genetiques dans un reseau
CN111224905A (zh) 一种大规模物联网中基于卷积残差网络的多用户检测方法
Huang et al. Design of learning engine based on support vector machine in cognitive radio
CN110808932B (zh) 基于多分布测试数据融合的多层感知器快速调制识别方法
Yu et al. Deep reinforcement learning for wireless networks
Wu et al. FedProf: Optimizing federated learning with dynamic data profiling
Celik et al. At the Dawn of Generative AI Era: A Tutorial-cum-Survey on New Frontiers in 6G Wireless Intelligence
Jalali Wireless channel equalization in digital communication systems
Amanna et al. Hybrid experiential-heuristic cognitive radio engine architecture and implementation
Li et al. Dynamic multi-channel access in wireless system with deep reinforcement learning

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
NENP Non-entry into the national phase in:

Ref country code: DE

WWW Wipo information: withdrawn in national office

Country of ref document: DE

122 Ep: pct application non-entry in european phase