Games of Daze

home *** CD-ROM | disk | FTP | other *** search

/ Games of Daze / Infomagic - Games of Daze (Summer 1995) (Disc 1 of 2).iso / x2ftp / msdos / ai / gaf200b / doc / gaf.doc < prev next >

Wrap

Text File | 1994-10-27 | 48KB | 1,009 lines

Introduction This document is intended to be the user guide for GA-FuzzyWare. It covers basic concepts of GAF (sections 1 - 4), interactive interfaces (sections 5 - 14), and command arguments and configuration (sections 15 - 16). Section 1: Running Modes A brief description of all running modes provided by GAF. Section 2: Schedule Policy Describes the scheduling policy under different schedule modes provided by GAF. Section 3: Emulation Covers essentials of the emulation segment(s) under GAF. Section 4: Adaptation The details about how adaptation works in the GAF environment. Section 5: System Menu Explains the menu bar of GAF, and how it works. Section 6: GAF Windows Explains different windows provided in GAF and covers contents of these windows. Section 7: File Menu Section 8: Run Menu Section 9: Option Menu Section 10: Segment Menu Section 11: Display Menu Section 12: Adapt Menu Section 13: Method Menu Detail descriptions on the use of all the pull down menus. Section 14: Function Keys Covers available function keys provided by GAF. Section 15: Command Syntax Explains how to invoke GAF from the DOS and available command arguments. Section 16: Configuration Brief describes the start up configuration file. GAF Conventions The following conventions are used throughout this document. 'Enter' means the Enter key or the Return key. Type 'key' means press the specified key without 'Enter'. Press F? means press the specified function key without 'Enter'. Enter means type in the specified text and followed by 'Enter' key. Select/Use an option: use cursor keys to move the highlighted option and then press 'Enter', or type the character on the left. GAF is not case sensitive, use either upper or lower cased key. Other GAF References Please refer to following GAF documents: fcl.doc fuzzy control language document tutorial.doc tutorial for construct a GAF segment demo.doc how to run the demon example Running Modes There are five running modes provided by GAF: control, monitor, check segment, simulation, and adapt. Note that control and monitor modes are not supported for current release. Control Mode The control mode provides users for online control and limited user interfaces for interacting with the real control environment and fuzzy rule segments. The output of all segments are routed through output channels, and the inputs are brought in through the input channels. Monitor Mode The monitor mode provides users for online monitoring and analyzing the real control environment and fuzzy rules. The output of all segments are for display only, they are NOT routed through output channels, but the inputs are brought in through the input channels. Check Segment This feature allows user to examine one segment at a time. Combined with other segment functions, such as preset and modify variables, user can check the response of the selected segment with fixed (constant) or manual values for inputs and local variables. Simulation Mode The simulation mode offers user to simulate the system with feedback or emulation segment(s) for integration test. The emulation can be user defined segment(s) or user provided data set with EMUL_SEGMENT or FEEDBACK keyword. The simulation environment allows user to tune up and debug control system with features such as single step and others. Please refer to Emulation section for more details about feedback segments. Adapt mode The adapt mode provides an offline adaptation of the fuzzy control system. An evaluation segment and its corresponding evaluation result score (an output variable) must be specified. When adapt mode is selected, GAF automatically optimizes the control segment by adjusting/adding/deleting rules and adjusting fuzzy membership sets. Please refer to the Adaptation section for more details of genetic- adaptation and also refer to FCL.DOC for defining an evaluation segment. Schedule Policy There are four different time modes: real time mode, simulated time mode, sim-schedule mode, and 24 hours clock mode. The real time mode uses actual system clock (computer clock) for scheduling the segments. This mode is the only allowed mode for running in the control mode. The following is an example of how real time mode schedules a three segments system with 0.1, 0.2, and 0.3 cycle time. Assume that each segment takes 0.05 second to run, and all three segments are enabled all the time. GAF scheduling table will be: Sweep # Time Seg 1 Seg 2 Seg 3 -------- ----- ----- ----- ----- 1 0.00 Run Run Run 2 0.15 Run --- --- 3 0.20 --- Run --- 4 0.25 Run --- --- 5 0.30 --- --- Run Instead of using real system time, the simulated time mode uses fixed time interval to simulate the schedule rate for the system. For example if the user specifies schedule interval is 0.1 second. Then the above scheduling table will be: Sweep # Time Seg 1 Seg 2 Seg 3 -------- ----- ----- ----- ----- 1 0.0 Run Run Run 2 0.1 Run --- --- 3 0.2 Run Run --- 4 0.3 Run --- Run 5 0.4 Run Run --- The sim-schedule mode resembles the simulated time mode, except that it does not use the cycle time of each segment for scheduling. Instead the sim-schedule mode uses the schedule rate for all segments, i.e. runs all segments all the time. Therefor, the scheduling table will be: Sweep # Time Seg 1 Seg 2 Seg 3 -------- ----- ----- ----- ----- 1 0.0 Run Run Run 2 0.1 Run Run Run 3 0.2 Run Run Run 4 0.3 Run Run Run 5 0.4 Run Run Run The simulated clock mode is a variation of sim-schedule mode. This mode allows user to simulate 24 hours clock within a single plot frame. Emulation An emulation segment is just like a control segment with EMUL_SEGMENT or FEEDBACK keyword instead of SEGMENT. You can define the feedback segment with fuzzy rules, math formulas, or data tables. Multiple feedback segments are allowed. In simulation mode GAF automatically closes the control and feedback loop with the same variable names. For example, if there is an input variable "POS_ERR" defined in the control segment, and it is also declared as an output in the feedback segment. Then GAF automatically uses the output value from the feedback segment as the input value of the control segment. On initialization, GAF searches through all segments to find matches among regular (control) and feedback segments. Please refer to Fuzzy Control Language (FCL) guide for constructing a GAF segment. GAF supports two kinds of feedback modules. The first one is using user data, from field measured data or generated from higher level model. The other choice is to use user defined feedback segment. To use user data as feedback, user simply put data in a text file with spread sheet like format, i.e., one column per variable, with different variables in a row. The data set must be an evenly distributed complete data set with both input and output variables declared as a normal segment. GAF only has limited capability to handle non-evenly distributed data set. The output variables defined in this file will be treated as output of a feedback segment. Alternatively, user can use the combination of fuzzy rules and math processing (pre and post processing) to build an emulation segment to close the feedback loop in order to simulate the system. One of the important features in GAF is that you can tune your system by trying different schedule rate (or cycle time defined in each segment). There are different time modes offered by simulation mode and check mode. These time modes are: real time mode, simulated time mode, sim-schedule mode, and 24 hours clock mode. Please refer to the Schedule Policy section for details about these time modes. Adaptation While fuzzy logic mimics the human's imprecise reasoning, the genetic algorithm mimics the evolution of the nature. The genetic algorithm uses randomization techniques to reach local optimum without exhausting searching through the solution state space. With randomization, genetic algorithm is also capable of escaping from local optimum and hopping into a new better state. GAF uses genetic algorithm to derive proper rules and fuzzy sets from the initial rules. By changing, adding, deleting rules and fuzzy membership sets of the fuzzy system, the genetic algorithm automatically adapts and optimizes the fuzzy control system. Besides the feedback segment, the adapt mode also needs an evaluation segment for evaluating the result of adaptation. With current release, user must build an evaluation segment and specify the result variable of evaluation. A canned evaluation will be supported in future release. Basics for Adaptation The basic concept of genetic-adaptation process is that GAF produces a new segment (by mutation or crossover from existing segment(s)) and then evaluates the new segment and compares its result against old ones (saved in the population pool). If a better result is obtained, the new segment is inserted into the population pool and the worst one in the pool will be removed. After starting adapt, GAF evaluates the selected adapt segment and saves it on the top of the population pool. The pool can save up to one hundred best generations in the order of their performance -- the evaluation result. GAF then repeats the genetic-adaptation process -- creation, evaluation, and judging the result, until reaches a pre-specified maximum iteration or quit from the user. Reproduction from Elites One of the concept in genetic algorithm is that the fittest gets the highest chances to produce its off-springs, so it has the best chance to preserve its better genes. GAF uses weighed random to select one or two segments out of the population pool to reproduce a "next generation" based on one of the methods described under Adaptation Method. So, the fittest generation, which has the best evaluation result, has the highest weight in the weighed random selection. Evaluation Criteria The evaluation process for genetic-adapt is defined as follow. GAF runs through all preset cycles, which are preset conditions defined in evaluation and other segments. After each cycle the result score from the evaluation segment (see Evaluation under Option menu for defining result variable) is compared to the minimum score. If it is below the minimum and the population pool has been filled up, GAF stops further evaluation and delete this new generation. Otherwise GAF keeps evaluation through all preset cycles. The overall evaluation result will be the average score of all the preset cycles. User can specify the test time for preset test cycle and number of samples should be taken for evaluation. The test time is used to run each preset test, then GAF samples the score result for the specified number of times. The preset evaluation result will be the average of all samples, and the final result will be the average of all preset cycles. For example, if the test time is two seconds and the number of samples is 4 (with 0.1 sample rate), the total time for a preset cycle will be 2.3 seconds. Adaptation Method GAF uses weighed random to select one of the genetic algorithms for creating next generation. GAF allows user to change all predefined methods used by GAF. Followings are brief descriptions of these methods. Change rules The change rules method adapts the existing rules, i.e., modify existing rules with mutation and other algorithms. No new rule will be added and no rule will be disabled. Adjust input set The adjust input set method alters the fuzzy membership sets defined in the segment. The following algorithms are used to modify the membership set. Mutate The mutate method randomly change one or two "cell" in the membership set. Crossover The crossover method uses two parent genes (two generations from the gene pool) to generate next generation by inheriting cells from either parent. Intensify/dilute The intensify method intensifies the membership set, the dilute method dilutes the membership set. Broaden/restrict This method either broaden or restrict the fuzzy membership set. Shift fuzzy set The shift method shifts the membership set along the value without changing its shape. Change truth This method alters the truth value of the selected fuzzy membership set. Adding new rule This method creates new generation by adding new rule to the segment. The new rule is created with the same way as described in change rule. Enable and disable rule This method disable (or enable) rule to form a new segment. Alter cycle time This method creates new generation by modifying the cycle time of the old segment. The range of cycle time can be specified with max and min under adapt menu. Alter output gain This method adjusts the output gain to form a new generation. System Menu The GAF system menu bar, at the top of the screen, is the primary user interface to the system. The system menu bar contains following menu items: File, Run, Display, Option, Adapt, Method, Segment, and Help. Use the first character of each menu item or use cursor keys (right/left arrows) to select the menu item. When a menu item is selected a pull down menu appears to provide function selection. To select a function item in a pull down menu either by the character on the left of that item or move the cursor with up/down arrow keys and then "Enter". When a function item is selected, further secondary menu selection or data entry dialog may be provided by that function. Except rule selection, all secondary menus and data entry dialog boxes are displayed immediately beneath the pull down menu. GAF uses rule text window for rule selection. Press ESC in pull down menu and secondary menu will return to system menu (the menu bar). Press ESC under system menu will bring up the Run menu, except when the system menu was ESCaped from any of the running mode (i.e. from adapting or simulation). Before discussing each menu item in detail, the screen windows are explained for clarification. GAF Windows There are five types of system windows and four types of display windows. The system windows are: menu window, hint window, adapt status window, segment title window, and segment display window. The segment display window contains four sub-windows: rule text window (text window), trend plot window (plot window), combined result window (result window), and fuzzy rule firing window (firing window). Menu Window The menu window is at the top of the screen occupied by the system menu bar as described above. In any of the run mode, the menu window is closed; use ESC or F10 to get back to system menu. Hint Window The hint window is at the bottom of the screen to provide user hints (available keys or functions) to the user. Note that in run modes the hint window also displays the current schedule rate. Segment Title Window The segment title window (title window) shows the information for the selected segment, see Segment section for selecting a segment. The title window is located immediately above segment display window. Example of title window: Segment 0 20 Symbols 2 Inp VAR: MOVEMENT POS_ERR Cycle 0.06 11 Rules 1 Out VAR: CURRENT Where, the first line shows the segment ID, number of fuzzy membership symbols, number of input variables, and the names of input variables. The second line shows the cycle rate of the segment, number of rules, number of output variables, and the names of output variables. Adapt Status Window The adapt status window (adapt window) shows the status of genetic adaptation. Adapt window only exists under genetic adapting mode, it is immediately above the title window. Example of adapt window: Iteration=50 Items=10 Best=0.95 Last=0.85 Worst=0.55 Good=11 Bad=39 Generation=4 From=2 Method=ADJUST_INPUT_SET.INTENSIFY_DILUTE Mutated set NEG_LARGE defined for POS_ERROR Average=0.935 Result= 0.935 0.935 0.935 0.935 Where: Iteration indicates number of iterations Items items in the best list (gene pool) Best the best score in the gene pool Last the last score in the gene pool Worst the worst score of all iterations Good number of good iteration Bad number of bad iteration Generation the generation of current iteration From which one (in the pool) been picked for adapting Method the method used to adapt Average average score for all tests Result score for each test The third line indicates what was changed for the new generation. In this example, the fuzzy membership set NEG_LARGE defined for variable POS_ERROR had been mutated (by INTENSIFY_DILUTE). Segment Display Window The segment display window (display window) displays important status and information of the selected segment. The display window occupies the most part of the screen immediately above the hint window. There are four types of sub-windows inside display window for showing fuzzy rules and running status: rule text window (text window), trend plot window (plot window), combined result window (result window), and fuzzy rule firing window (firing window). The place and area of these sub- windows can be rearranged by the user, please see Display menu section for rearranging display windows. Text Window The text window shows the rules and status of the rules for current selected segment. Each rule is presented in a row with several columns associated with it. The first column indicates whether the rule is disabled (marked with 'x') or fired (marked with '*'). The second column shows the rule number. The rest of columns display the content of the rule, i.e., the fuzzy membership set symbol for corresponding inputs and outputs. These content columns are presented, in different color, with inputs first then the outputs as the same sequence shown in the title window above it. If local variables are selected (see Display section) the local variables are displayed after input variables. In run modes, use cursor keys to scroll up, down, right, and left for undisplayed contents. Trend Plot Window The plot window draws the trend graph for all inputs, outputs, and selected local variables with time in x-axis and value in y-axis. Each variable has its own sub-window, started from top to bottom, with the same sequence and color as in the text window. See sub-section "Set Plot Time Frame" under Display for changing the frame duration. Combined Result Window The result window depicts the result for selected output of the segment. The trapezoid result of all fired rules, which generate non- zero output, are in the upper portion of the window . In the lower portion of the window the defuzzified result (combined result) and the output value are displayed. Multiple outputs can be selected with split display area. See sections Segment and Display for selecting output and arranging display window. Rule Firing Window The firing window shows the fuzzy inference status (firing condition) of selected rules. This feature along with the stepping mode, under simulation or check segment, is the best way to tuneup or debug the control system. You'll find the selected rule's fuzzy sets are displayed to the left of the combined result. The corresponding input value for each set is plotted vertically in the same window. Note that each fuzzy set input has the same color as in the text window. Refer to Display section for arranging display window, and Segment section for selecting rules. File Menu: File Functions The File menu provides the following functions: Q: Quit Select 'Q' or use ALT-Q to exit the program. O: Open segment Select this option to open a segment from the specified file, GAF will prompt for the name of the file. If the default directory is specified (command switch /DIR) and the entered file name does not contain path, the default directory will be searched for the specified file name. The segment file can also be specified as the command arguments. When open a new file segment a sequence number will be assigned to that segment as its segment ID. The segment ID is used for displaying segment title. C: Close segment Use to close a segment. When selected, GAF pops secondary menu for a list of all segments for user to select appropriate segment to close. Note that a closed segment will be marked as "disabled" for current release, i.e., the memory will not be released. B: Report selected best list A: Report all items in list R: Report file These options allow user to generate report about the item(s) in the best list (population pool). Use 'R' to set the file name of the report, GAF prompts for the file name. Use 'A' to report all items in the list, and use 'B' to select one out of the best list. When 'B' is selected, GAF will prompt for the item number to be reported on. The item number must be between 0 and less than the current population in the pool. L: Enable adapt log D: Enable detail log G: Log file to In adapt mode, GAF can generate a log file for user to review. Use 'L' to enable (disable) the log, use 'D' to switch on (off) detail log. Use 'G' to specify the log file name. E: Enable plot data P: Plot data to When enabled, GAF generates a data file for the result of each iteration. The data file can be used by MyPlot to plot. The 'P' option specifies the file name of the data file. S: Generate statistic log T: Statistic to These options allow user to generate statistic report about the adaptation. Use 'T' to set the file name of the statistic log. The log includes statistics for all methods. Based on the statistics you may want to adjust the weight of the adapt methods. Run Menu: System Run Modes The Run menu provides different modes for running GAF system. As mentioned, GAF offers following five running modes: control, monitor, check segment, simulation, and adapt. Also provided is to terminate (quit) the program. Although there are five running modes, they can be divided into two categories: the control (simulation) portion and the automatic adaptation. While in any of the running modes, use ESC or F10 to escape back to or bring up the system menu. After bring up the system menu other functions provided by pull down menus are available except specifically noted. Upon completion of other functions, ESC again will bring you back to original running mode. Use Alt-Q to quit the program. Q: Quit the program Select this option or type 'Q' to quit the program. R: Control Mode M: Monitor Mode These two modes are disabled for current release. C: Check Segment Use this option to start checking the segment specified from the Segment menu. Use Segment menu to select proper segment, then start the check mode by selecting this option. Basically, the check mode is the same as simulation mode, except that only the selected segment is enabled. For available options during checking, please see simulation mode's user interfaces. Simulation Mode Use this option to start the simulation. Note that you must have feedback or emulation segment(s) with EMUL_SEGMENT or FEEDBACK keyword. Or use the Option menu to specify the feedback segment. When the mode is selected, the current selected segment will be displayed if it hasn't already been. Use any of these keys, described as follow, to start the simulation. Under monitor, check segment, and simulation modes, the following interactive user interfaces are supported. C : toggles on/off continuous mode F : continue until the end of the plot frame 1 ~ 9 : run 1 ~ 9 steps 0 : run 10 steps + - : increase/decrease sample rate (see below) enter : single step ESC : go to system menu bar alt-Q : quit the program F1 ~ F9: see function keys cursor keys to scroll up, down, right, and left of text window All other keys, while not in continuous mode, are treated as "Enter" for single step. Note that during continuous mode only the following keys would halt the continuous process: C, ESC, and function keys. The cursor keys will not halt continuous mode, instead, GAF performs the scroll and continue its simulation. There is an easy way to change the sampling rate in sim-schedule and simulated time mode. That is to use the '+' and '-' keys. By press '+' or '-' key, GAF adjusts the schedule rate or cycle time by increasing or decreasing an adjustable amount. Please refer the Option menu section for setting the adjust amount. The current sample rate value will be displayed in the hint window at the lower right corner of the screen. A: Adapt mode When adapt mode is selected, GAF starts adaptation of the selected control segment. The followings must be specified before starting adaptation: the segment to be adapted (optimized), the evaluation segment, and the variable for the evaluation result. GAF stops if any of the required is missing. The following interactive interfaces are supported in adapt mode. C : continue Alt-C : switch to continuous mode Alt-P : turn on iteration pause Alt-A : turn on preset cycle pause Alt-Q : quit the program ESC : go to system menu bar F1 ~ F6: see function keys cursor keys to scroll up, down, right, and left of text window All other keys will pause adapting temporary. Option menu: Change Parameters The option pull down menu allows user to change the simulation and evaluation environment. Currently only following functions are supported. R: Real time mode T: Sim time mode S: Sim schedule mode K: Sim clock mode These four modes are the basic scheduling policy described in the Schedule Policy section. The selected mode will be marked with "x" on the right. C: Sim schedule rate Use this option to set desired time interval in seconds for either simulated time mode or sim-schedule mode. Current value id displayed on its right. I: Increment value As mentioned, an easy way to change the simulated schedule rate is to use the '+' and '-' keys to increase and decrease the schedule rate. This option allows user to set the amount of increasing/decreasing, the default value is 0.005 second. E: Emulation segment This option is used to specify the feedback segment. Normally, this option is only used to show the feedback segment, if proper key word (FEEDBACK or EMUL_SEGMENT) is used to define the emulation/feedback segment. When selected, a secondary menu contains all segment names are displayed for further selection. V: Eval segment This option can be used to specify a user defined evaluation segment instead using EVAL_SEGMENT key word to declare it. The current evaluation segment ID and its name are shown, if it is specified. When selected, a secondary menu pops up with names of all segments to allow user to select the proper segment. A: Eval result variable Use this option to specify the evaluation result variable. When selected, a secondary menu pops up with names of all variables of the specified evaluation segment to allow user to select the proper variable. GAF uses "EVAL_RESULT" as default variable. B: Eval null band This option allows user to specify a null band of evaluation. The null band is used by GAF to determine whether it's necessary to check there are identical generations or not. Segment Menu: Change segment and value The segment menu offers user three major functions: Select a segment for display or for following actions Change/modify the segment Control/change the display of the segment S: Segment number Use this option to select a segment. The default selected segment is the first segment (the segment 0). The selected segment is displayed in the display window and title window. When selected, a secondary menu pops up with names of all segments to allow user to select the proper segment. E: Segment enabled To toggle enable or disable the selected segment. If the segment is enabled it will be marked with "x". Note that when a segment is disabled with this option, it will not be enabled by the "ENABLE" condition defined in the FCL, however, when the option is enabled then the "ENABLE" condition determines whether the segment is enabled or not. C: Cycle time Change the cycle time for the selected segment. Note that the real cycle time may depend on the running mode and time mode been selected. Refer to Schedule Policy section for more information. I: Set input value O: Set output value L: Set local variable value These options allow user to manually set a variable's value. After select one of these options, GAF pops up a secondary menu with all variable names alone with their variable number. Select a variable, GAF then requests for new value. Enter the new desired value will close all menus. Note that you can use F7 to initialize the system, see function key for detail. G: Set output gain Use this selection to change the gain of selected output. Again, just like setting variable value, GAF asks for output variable and then the new gain value. N: Enable rule D: Disable rule GAF allows you to enable or disable the rule(s) in any of the running modes. The disabled rules are marked with 'x' at the first column in text window. Select one of these options, GAF then highlights the first rule at the second column in text window. Use cursor keys to move highlighted cursor and then press 'Enter' to enable or disable the rule. A: Enable all rules This option gives user an easy way to enable all rules. R: Display result After changing the Display window's arrangement, i.e. selecting whole display or no plot mode, use this option to select the desired output you want to display. Note that GAF automatically selects the first output for display, if none is selected. F: Display rule firing The above option (Display result) allows user to specify the output to display, this option allows user to select rule to be displayed with its detail inferencing in firing window. When select this option, like disable/enable rule, GAF uses the text window for selecting menu. Note that selected rules are marked with 'x'. M: Combined result only When this option is selected, GAF deselects all rules from displaying its inference (firing window). So only the combined result will be displayed. P: Preset to Use this option to preset to a specific state defined by "PRESET" statements in segment(s). GAF will prompt for the preset number. Please refer to F8 of function key for similar function. Display Menu: Arrange Display Windows The display menu allows user to arrange and select the screen layout, text font, and color of the screen display. 1: Use standard font 2: Use small font With these two options, you can switch between normal text display or small font text display. 3: Select normal display mode The normal display mode displays the text window on the left portion of the display window and the plot window on the right. See window split for changing the size of text window. 4: Select whole display mode In this mode the text window still occupies the left portion, but the right portion of the display window splits into two parts, the upper half shows the plot window, and the lower half displays the combined result and/or firing window. See Segment section for more information about selecting rules and outputs. See window split for changing size of text window. 5: Select plot only mode This option allows you to see only the plot. GAF dedicates the entire display window to plot sub-window except reduces the text window to only first two columns on the left, which shows the status and rule number. 6: Select no plot mode This display mode removes plot window from the right portion. So the combined result and/or firing window are displayed on the right along with text window on the left. 7: Select text only mode This mode shows the text window only. Because graphics, i.e. trend plot, combined result, and firing condition, takes much of CPU time, this mode allows you to speed up the test or control. P: Plot duration The default plot frame period is five seconds (except in auto clock mode). Use this option to change the frame duration, the value is specified in seconds. C: Color/monochrome Use this option to switch between color and mono mode. W: Window split (Change size of text window) Allows user to change the size of the text window. The value is specified in percent, i.e. 60 means that text window occupies 60% of the display window. If the value is greater than 90, GAF automatically switches to text only mode. If the value is less than 10, GAF shuts text window off (except the left most two columns). D: Display local variable The default display mode is to display/plot IN, OUT, and INOUT variables. The LOCAL variables are not displayed. Use this option to enable LOCAL variables. Adapt Menu: Setup adapt environment The adapt menu is used to change the overall environment for genetic adaptation. See the Method section for more information about controlling the genetic algorithm. C: Continuous mode Set the adaptation to continuous mode without pause or delay, unless user interferes. The continuous mode is the default mode after starting adaptation. P: Iteration pause Y: Pause per cycle These two options allow user to halt the adapting process in order to examine it. The iteration pause is to halt at the end of each iteration. The cycle pause is to halt at the end of each preset cycle. Type 'C' to continue, press Alt-C to switch back to continuous mode. Also see Iteration and Cycle delay. S: Sample number Specifies number of samples should be taken for calculating each preset test's evaluation result. W: Weight increment GAF uses linear proportional weight for random selecting from the gene pool. This option is to set the linear increment of the weight. For example the increment is 2, then the weight for the best in the gene pool will be 2 * 100 = 200, and the weight for the worst in the pool will be 2. B: Max best items The number of items kept in the best list (gene pool). The limit of this value is 1 to 10. I: Max iteration This allows user to specify the maximum number of iterations. GAF stops after this number of iterations is reached. If the value is less than or equal to 0, there is no limit. T: Test time As explained the test time is the time for each preset test before evaluate the result. Use this to set the proper value, in general it should be greater than the rising time for the longest response in all preset tests. D: Iteration delay L: Cycle delay Instead pause for each iteration/preset cycle, users can specify time delay between iterations or preset cycles. Zero value means continuous. O: Minimum score This value defines the minimum score for a successful preset test cycle. GAF tries to go through all preset tests, but stops after any preset test failed to score the minimum value. The default value is 0.6. A: Adapt segment Use this option to change the segment you want to adapt. The name of the current selected adapt segment is shown on the next line in the menu. When selected, a secondary menu pops up with names of all segments to allow user to select the proper segment. Note that the selected adapt segment may not be the same as the selected display segment. U: Use best list G: Use original segment The first option allows user to pick one adaptation out of the best list and then switch to simulation mode for testing or debugging. The "Use original segment" option allows user to switch back to the original segment, which is loaded from the file. R: Reset best list This option allows user to clear the gene pool, i.e. delete every sample in the gene pool. Method Menu: Control Genetic Algorithm The method menu allows user to control the adaptation algorithms by setting new weight factor. A method can be disabled by setting zero weight. For example, if no new rules are allowed then set "adding rules" to zero. The method menu is divided into two parts, the upper half for changing weights for overall adapting methods. The lower half is to change the detailed weight for specific adapting method. Z: Zero all weights This option zeros all weights in the upper half adapting methods. C: Change rules The option changes the overall weight for adapting existing rules. I: Adjust input set 4: Adjust input set ... The first option (in the upper half) changes the overall weight for adapting input fuzzy membership set. The second option is used to change the detail adapting weights inside input set adaptation. When the second option is selected, GAF pops up a sub menu under method menu. The followings are options in the sub menu. M: Mutate Use this option to set the weight of mutation. C: Crossover To set the weight of crossover for adjusting fuzzy set. I: Intensify/dilute Set the weight of intensify and dilute for adjusting fuzzy input set. B: Broaden/restrict Set the weight of broaden and restrict for adjusting fuzzy input set. S: Shift fuzzy set Set the weight of shift fuzzy input set. T: Change truth Set the weight of adjusting the truth of an input set. K: Scale (0 - 1) The scale factor defines how much change is allowed for the above methods, except mutate and crossover. The amount is the percentage of the member set. One means 100% and zero means none. A: Adding new rule This option sets the weight of adding new rules into the adapting segment. See Maximum rules to define the limit of rules. E: En/dis-able rule This option sets the weight of enable/disable existing rules in the adapting segment. T: Alter cycle time This option sets the overall weight of altering cycle time of the segment. The range of cycle time can be specified with max and min cycle time. G: Alter output gain This option sets the weight of adjusting the output gain to form a new generation. 1: Max cycle time 2: Min cycle time These two values clamp the sample rate, the cycle time of the segment. GAF will not exceed these limits during adaptation when adapting of cycle time is enabled (the weight for alter cycle time is greater than 0). 3: Maximum rules This option allows user to set a limit for number of rules is allowed during adaptation. If the number is less than 1 there will be no limit. Function Keys F1: Get help GAF provides context sensitive help. Press F1 to call up the on-line help. Inside help you can jump to other topic by selecting listed topic. F2: Display the system info Press F2 to see the system limits. The system info shows the version of GAF, system wide limits, and current free memory status. IO length (in bits) the supported analog IO bits Number of sets per segment maximum number of fuzzy input sets in a segment Number of rules per segment maximum number of rules in a segment Total variables per segment maximum number of variables (IN, OUT, and LOCAL) in a segment Number of segments maximum number of segments in the system Number of fuzzy symbols maximum number of fuzzy symbols in a segment Name length (IO & symbol) number of characters for names (variable name or symbol name) Number of best genes the maximum limit of the gene pool (best list) F3: Next segment Press F3 to switch the display segment, i.e., cycle through all segments. F4: View segment information Press F4 to see the variable definition, initialization and run functions, and other information for the selected display segment. F5: View symbol definition Press F5 to see the fuzzy set symbols' definition for this segment. F6: View preset definition Press F6 to see the preset definition for this segment. F7: Initialize the system The F7 function key allows you to initialize the system. This is defined in each segment's INITIALIZATION declaration. GAF goes through all segments to execute the initialization defined in each segment and then resets the segment scheduler. F8: Preset the system The F8 function key allows users cycle through all presets defined for all segments during simulation. F9: Zero the system Press F9 to zero all the inputs and outputs for all the segments during simulation. Command Syntax The command syntax for GA FuzzyWare is: GAF [switches] [segment files] Available command switches are: /DIR=dir_path to specify the default directory for segment files /CFG=file_name to specify the configuration file /H or /? help Examples: GAF /dir=c:\gaf a1.seg a2.seg a3.seg specifies all a1 ~ a3 files are under c:\gaf directory. GAF /dir=c:\gaf app.s00 /dir=c:\app app.s01 app.s02 specifies app.s00 under c:\gaf, both app.s01 and app.s02 are under c:\app directory. Configuration When starts up, GAF searches configuration file in following order specified by /CFG command switch "GAF.CFG" under current directory "GAF.CFG" at where GAF.EXE resides The configuration file is a text file for setting up the GAF environment. The format of the configuration file is a series of "attribute = attribute_value;". The attribute can be in any order and can be defined multiple times with the latter value overrides the previous value. The configuration attribute can also be included inside FCL module file by enclosing it within "CONFIGURATION" and "END_CONFIGURATION;" keywords. Because FCL modules are processed after GAF.CFG, so FCL's configuration will override GAF.CFG. Please refer to FCL.DOC for details.