Patent Application
20060256962
This invention relates to random number distribution generation and more particularly to systems and methods for producing random number distributions in devices having limited processing and storage capabilities.
The usual approach to generating random number distributions is to transform one or more standard uniformly distributed independent random numbers into one or more random variables with a given distribution. Many of these rely on heavy amounts of computational ability, such as is available on a PC and are thus not practical for use in devices, such as cell phones or other small relatively portable devices.
Random number generators can also be created in a purely hardware environment thereby bypassing the need for large storage or CPU capability. A particular type of digital electronic circuit known as a Linear Feedback Shift Register (LFSR) can be used to create a range of “random” values called Pseudo Random Binary Sequences, (PRBS). The properties of these circuits are well known, and their direct output is a Binomial Distribution, the discrete time equivalent of the Poisson Distribution. However, in applications where specific, and possible changing, distribution shaping is required, PRBS's are not adequate. For example, in situations with limited processing power, i.e. pure hardware digital circuitry with little or no access to a microprocessors, the range of random numbers available for use can be severely limited, both in the type of the distribution, and the range of statistical properties within the same distribution type, i.e. minimum, maximum, mean and standard distribution. The available “length” of any sequence may be severely limited before the sequence starts repeating from the beginning. This may lead to situations where the randomness is violated for decent results in the application of interest.
Arbitrary numerical distributions are presented for use in devices having limited processing and storage capabilities by having the device accept strings of arbitrarily distributed numbers from a source outside of the device. In one embodiment, a master controller creates a table of values which follow the desired minimum, maximum, mean, and standard deviation etc. of the particular desired statistical distribution required. The created table is then communicated to the limited capacity device and can be used whenever a distribution of random values is required. The master controller could have one of several slave devices associated with it in the system.
In another embodiment, where the storage capability of the device is large enough to store a table of values with sufficient different entries to create a distribution of satisfactory “randomness” for the particular application, a random number generator within the device is used to select the order of presentation of the table of values.
Devices 20 and 30 are but two examples of embodiments utilizing the concepts of the invention in which number sequences having the desired statistical distribution for a given period of time are remotely generated. The remote generator has resources to produce high quality numbers, such as statistical distributions, copies of actual events, simulation of actual event according to a distribution, an exponential distribution or simply act as a high quality random number generator. These number sequences are transmitted from the source of generation (which can be a PC or any other source), through a communication network such, as communication network 101 to the limited resource device. Network 101 could be wireline (including media such as fiber), wireless or combinations thereof.
The LFSR is a good choice because it works well on a limited processing environment, but other circuits and methods could be used to generate the necessary randomness. Load control 21 gives a generate instruction to LFSR 24 which then generates a random number. When each input of the LFSR flip-flops are loaded with a seed value (anything except all 0s, which would cause the LFSR to produce all 0 patterns) all that need happen is for the LFSR to be clocked and it will generate a pseudorandom pattern of 1s and 0s. Note that the only signal necessary to generate the number of pattern is the clock. The pseudo random number from LFSR 24 is used to address one line of the previously stored data. This addressed data, which initially was generated by device 11 (
In one embodiment, the circuitry 25 could be a counter which is set to a number and counts to zero. At zero it produces an output signal. The random number(s) from RAM 22 could be used to provide the “set” point for the counter such that the output pulse (when it is generated) represents the time between certain events. The pattern of the true events that this circuit is trying to mimic was established by remote generator 11 because device 20 does not have the capacity to generate the proper number patterns without external help. The signal from circuitry 25 also is used to trigger load control 21 such that either another pseudorandom address is generated by LFSR 24 to access the data already stored in RAM 22 or to control the storing of new data from remote generator 11.
Note that, if a system has a plurality of limited resource devices 20, each such device could be configured with different initial values for the LFSR so that the output from each device 20 would be different even though they each used the same number sequences from remote generator 11. Also note that the LFSR need only be of sufficient size to create an address bus the width of the RAM address bus. If sufficient resources are available, the output of the LFSR could be a longer address then is necessary, with the address being cut into slices to be used for different cycles if desired. This would allow a table (a download of number sequences) having, for example, 256 entries to be accessed by taking eight address bits from a ten bit LSFR. That would mean that the table would be fully accessed four times for one complete cycle of the LSFR, but the accessing of any one address would be scattered across four access times as opposed to one. Thus, full access and then entry N could be picked four times while entry N+1 may only be picked two times, depending on the method of slicing. This system makes it even more random than using eight bits from the LFSR to select and eight bit address which would produce the same sequence four times in 1024 “clocks.”
In certain embodiments the RAM could have a default table so that on power up (or other loss of input data) LFSR 24 could access randomly one of the numerical sequences already stored in RAM 22 while it is waiting for a message from remote generation 11. This loaded data could, for example, be utilized in situations where for bandwidth or other reasons the data is not moved into RAM 22 fast enough or in an interrupted fashion from communication network 101.
The data stored in RAM 22 can be data from a previous download or data provided from a memory device permanently installed in device 20. Whenever it is desired to change the number sequences or to use different random distribution patterns, remote generation 11 changes its pattern of data which is then stored in RAM 22 and the cycle repeats.
Note that as long as the average rate of consumption of the user (device 33) is balanced by the average rate of arrival of new values in messages from remote generation 11, the size of the FIFO need only be large enough to hold the entries in one message give or take a few entries for variable delay in a communication network. Since remote generation 11 has much higher processing and storage capacity than does device 30, much larger sequences of random values are possible than would be possible if the data were to be generated in device 30.
FIFO 31 could be modified to have default values available for power on which value could be used until FIFO 31 receives additional number sequences from remote generation 11. In addition, if a delay or break in communication occurs such that FIFO 31 does not receive additional values for a period of time, FIFO control 32 could operate to effectively move the pointers of FIFO 31 to reuse the last data set (message) thereby allowing values to continue to be fed to circuit 33 albeit with a repetition in the distribution being created for a period of time until communications with remote generation 11 is reestablished.
Note that devices 20 and 30 can be combined depending upon whether it is desired to use the random memory buffer or simply allow the remotely generated number sequence to flow through a FIFO. As discussed above, the circuitry of devices 20 and/or 30 can be constructed using ASIC or FPGA technology or any other technology as desired, and the number sequences in remote generator 11 can be computer generated, circuit generated, or can be generated as a result of the monitoring of actual test results.
Also, note that if bandwidth becomes an important issue at any period of time, each of devices 20, 30 has its own method and system for coping with the inability to continue functioning for a particular time. This is accomplished with respect to
Although the present invention and advantages thereof have been described in detail, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims. Moreover, the scope of the present application is not intended to be limited to the particular embodiments of the process, machine, manufacture, composition of matter, means, methods and steps described in the specification. As one of ordinary skill in the art will readily appreciate from the disclosure of the present invention, processes, machines, manufacture, compositions of matter, means, methods, or steps, presently existing or later to be developed that perform substantially the same function or achieve substantially the same result as the corresponding embodiments described herein may be utilized according to the present invention. Accordingly, the appended claims are intended to include within their scope such processes, machines, manufacture, compositions of matter, means, methods, or steps.
