ReCiPe

This is a repository for tooling to support the ReCiPe formalism for modelling and verification of reconfigurable multi-agent systems.

Currently it contains the R-CHECK tool that includes a language to specify reconfigurable multi-agent systems, and support for the model checking of these through the state-of-the-art model checker nuXmv, described better below.

R-CHECK

Language

The language allows the definition of a set of agents that interact synchronously. See `./example.rcp' for an example.

LTL Model Checking

For model checking add specifications at the end of the script in the form of LTLSPEC <spec>;, where <spec> respects nuXmv syntax for LTL predicates. When making reference to the variables of an agent instance prefix them by the name of the agent in the following form id-var, similarly for their state: id-state, and any transition label.

To check whether the system is stuck, the predicate keep-all can be queried, e.g. F keep-all will be true if the system eventually settles in a state (where from then on no variable values may change).

When the agents use infinite-state variables the tool will use symbolic bounded model checking (using msat).

Simulation

Simulation is performed using nuXmv, with a symbolic model.

Simulation will show changes in variables in the next state only, with the exception of receive label variables which are only shown when they are true. Send labels do not appear in the simulation output since they are defined as nuXmv predicate definitions rather than as variables.

Compilation

This project was compiled and tested successfully with:

1. Java 15
2. nuXmv 2.0.0
3. Maven 4
4. NPM (to use frontend)

We suggest using the same or similar versions for compilation and running. Check pom.xml for the right versions.

For compilation to a single jar file simple run mvn clean package, and the jar file target/recipe-0.1.jar will be produced.

Execution

We assume that the following commands are available from the environment PATH variable:

1. nuxmv (for analysis and simulation)
2. `npm' (for frontend)

Or for the case of nuxmv:

The folder you downloaded from nuXmv developer site should be renamed nuxmv be placed in ./nuxmv/bin/

Note also that the executable /bin/nuXmv should be renamed in small letters /bin/nuxmv to avoid any interruption due to case sensitivity.

For windows users: make sure that you can access the localhost and that your firewall does not block that. This is important if you want to use the GUI app, but in any case, the command line should work normally.

After compilation, run the jar, e.g.java -jar ./target/rcheck-0.1.jar with the below arguments for the CLI functionality:

 -i,--input <arg>   info: input recipe script file
 -d,--dot           info: output agents DOT files
 -n,--smv           info: output to smv file
                    args: <recipe script>
 -mc,--mc           info: model checks input script file
 -bmc,--bmc <arg>   info: bounded model checks input script file
                    args: bound (by default 10)
 -sim,--simulate    info: opens file in simulation mode

and with the following to run the GUI app:

 -g,--gui           info: opens gui, (make sure 'npm install' is called in ./frontend-react before, and that ports 54044 and 3000 are free)

For example, when you are in the main folder "rcheck" of the artefact submission, you can do the following:

Go to https://nuxmv.fbk.eu and download nuXmv for command line of your operating system. You will get a folder and you should name it nuXmv and place it in the main folder ./rcheck

Now you can execute the following commands to work with the GUI web interface for the first time, which will be served on http://localhost:3000:

 cd ./frontend-react
 npm install
 cd ..

Only execute the above commands once to build the web interface, and then you execute the following command to build the R-CHECK excutable:

mvn clean
mvn install

Then you can run the following to access R-CHECK's functionality through the web interface:

java -jar ./target/rcheck-0.1.jar -g

If you want to work directly with the command line tool, you have to supply the file input file with `-i' and specify what you want to do with it.

For the simulation from the command line, you can specify a constraint or you can just type "TRUE" to see a possible execution.

Example:

java -jar ./target/rcheck-0.1.jar -i example.rcp -sim