FMM Analysis Tool

There are two folders in this FMMFSM simulator: FMMFSM_analyzer, and FMMFSM_analyzer_Tl. There are three code files respectively for each folder.

• FMMFSM.py: Code for calculating the degree of membership of an FMMFSM.
• discreteSystemModel.py: Code for simulating the state transitions of a system model.
• outputChecker.py: Code for checking for mode confusions by comparing the FMMFSM and the system model. The usage and code is different for each directory as detailed below.

Please refer to the FMMFSM paper to more information about how to specify a FMMFSM and system model. Examples can be found in FMMFSM_tester/use_cases

FMMFSM_analyzer & FMMFSM_analyzer_Tl

FMMFSM_analyzer and FMMFSM_analyzer_Tl together make the FMMFSM analysis tool. FMMFSM_analyzer includes scripts that does not handle task labels, while FMMFSM_analyzer_Tl handle task labels.

FMMFSM_analyzer handles the random generatioin of action schedule. Please refer to the FMMFSM paper to more information about how to specify a FMMFSM and system model. Examples can be found in FMMFSM_tester/use_cases. Input for FMMFSM_analyzer should include the below information:

For FMMFSM, you need to specify:

• "initial_state_memberships": m_0 vector
• "input_fuzzified": alpha
• "transition_probabilities": phi

For system model, you need to specify:

• "initial_state": m_0
• "transitions"

Input for FMMFSM_analyzer_Tl is the same as for the FMMFSM_analyzer expect the user need to specify "task_labels" for both FMMFSM and system model. You could find an example in the FMMFSM_analyzer_Tl/config directory.

To use the FMMFSM analysis tool, you first run the run.py:

python run.py PATH_TO_FMMFSM_CONFIG_FILE PATH_TO_SYSTEM_CONFIG_FILE --num NUM_OF_ACTIONS --iter NUM_OF_ITERATIONS

PATH_TO_FMMFSM_CONFIG_FILE and PATH_TO_SYSTEM_CONFIG_FILE need to be in the FMMFSM_analyzer/config or FMMFSM_analyzer_Tl (if includes task labels) directory, and the corresponding output would be saved in the corresponding experiment folder (with time stamp) in the corresponding output folder.

NUM_OF_ACTIONS is the number of actions you want the simulator to randomly generate for a single file.

NUM_OF_ITERATIONS is the number of files you want the simulator to run for a single experiment.

Use --postp flag to save both the original results and the post-processed results (disregard all following errors identified after a blocking state).

Use --tl flag to use FMMFSM_analyzer_Tl instead of FMMFSM_analyzer.

An output from run.py includes:

• A computed directory, which includes one file for the FMMFSM simulation result (end with FMMFSM_Result.json) two files for he system simulation result (Sys_Binary.json: binary result; Sys_Readable.json: For readability), and a result file that includes the analysis result from outputChecker.py.
• A config directory, which includes all config files with the randomly generated action schedules.

After you run the run.py, you can run the analyze.py to summarize findings from the analysis results:

python analyze.py PATH_TO_EXPERIMENT_FOLDER

Use flag --postp to save the summarization for post-processed results.

Use flag --all to save the summarization for pre-processed results.

Use flag --save to save the summarization to a file in the output folder: we recommand you do this every time since the summarization could be too long to read in the terminal.

To test a FMM

You can apply -test flag to analyze a single pair of FMMFSM and system model. For FMMFSM, you need to specify:

• "initial_state_memberships": m_0 vector
• "input_fuzzified": alpha
• "transition_probabilities": phi
• "action_schedule"

For system model, you need to specify:

• "initial_state": m_0
• "transitions"
• "action_schedule"

"action_schedule" should be the same for a pair of FMMFSM and system model.