Matlab env dependencies:
- Matlab 2021b
- Simulink default
If you wish to use publicly available datasets, you can access third-party datasets. If you wish to use a random model, you can visit Dr. Shafiul's homepage. Get the latest experimental model generation tools SLFORGE/CYFUZZ
The format of the model is divided into mdl and slx according to the Simulink standard. If you use some third-party open source models, make sure that the dependencies between the models and the data are loaded into your memory. We do not recommend using third-party models in unfiltered conditions, which will cause your computer to crash abnormally. Please be careful
We are committed to testing the stability of the CPS (cyber physical systems) development tool Simulink. Our work mainly relies on the live mutation technique to detect Simulink bugs by means of Equivalent Modulo Input (EMI). We will first obtain the information of the test model, including the number of modules of the model, the coverage ratio, the input and output information of the module, the handle of the module and its connection, etc. The mutation testing step is performed based on the obtained model information. We add the conditional discriminant module to the random selection module, and synthesize discriminants with contradictory predicates according to the input and output information. Then another module in the model is selected according to our specific discriminant method to be added to the branch where the discriminant module is always executed. In branches that are not executed, we will synthesize new regions according to our method. This results in a variant model. Finally compare the variant with the original model to discover potential flaws in Simulink
Before starting, you need to set two environment variables SLSFCORPUS and COVEXPEXPLORE,both of which point to your reproduce/samplecorpus file. In subsequent operations, you need to put the test models into
this folder so that Simulink can quickly and accurately find the files you need
If you have questions about setting environment variables, please visit the official Mathworks website
In order to test the stability of the program, you need to test and collect the coverage information of the model in advance. You need to run this at the Matlab command line：
covexp.covcollect
You can change the collected configuration parameters in the file covcfg.m and set the maximum compile time for the model
Due to problems with some models, this process may experience program termination.
If you wish, you can choose in the file.Replace @emi.preprocess_models with @emi.check_SLforge . Also replace the path in the file with the absolute path where you store the model file
This will help you filter your models
Warning: You must collect coverage information before running the compiled test program.
You can modify the parameters of the mutation in the cfg.m file, including the number of variants, the rate of variants, and the choice of variant space.
Once everything is in place, you can run in the Matlab command line:
emi.go()
A mutation procedure for variants is performed.
The experimental results will be placed in the +emi_result file, and the information including the variants will also be written separately into the .mat data table of each
variant subfile. If you want to see the overall data, you can run from the Matlab command line:
emi.repoert
These errors in bug files can be reproduced with Matlab2021b
If there are two files in the folder that are original and equivalent, you usually need to open and run them separately. You can reproduce this error in the data comparator provided by Matlab. If the folder has only one file and the problem is not in accelerated mode or zero-crossing detection mode then you can just run and see the problem. If the problem with the folder is in acceleration mode or zero-crossing detection mode, in order to reproduce the problem you need to run the model separately in different modes. For example, the case where the zero-crossing detection model is wrong is the inconsistency between the heuristic zero-crossing detection algorithm and the zero-crossing detection algorithm disabled.
The last part referencing the model requires adding your folder to the Matlab path in order to run. Access to specific circumstances Mathworks website
05294630 Math Function error in accelerate by selceting Nan and zero
05296099 Data error for signal generator module in accelerated mode
05255309 Data Type error cause Cannot open or compile the file normally
05310042 Min module data inconsistency in accelerate simulation
05314520 The data type error after logging the signal is inconsisten
05314517 Reference model inheritance time exception
05274593 Abnormal If block condition judgment
05317645 Data type judgment error under reference model in Math operation module
05358090 Reference model sampling time inference exception
05358093 Heuristic inference exception in logging signal
05371387 zero-cross detection cause data exception
05320137 Exception output of abs block in accelerated compile mode
05382872 Compile error after logging signal in complex module
05382877 Accelerate simulation compile errors by using Lcc
05398645 Max module misbehaves under zero-crossing detection
05405356 Abnormal MAX zero-crossing detection in acceleration mode
Thanks to MathWorks consultants Zouyi Yang, Lily Yan and Finao Peng for their support. We got a lot of help from MathWorks staff in the discovery and confirmation of bugs, and we can't list them all. I would like to express my gratitude here.