rafdewes / hypermtl

Prototype Implementation of HyperMTL Monitoring. Auxiliary to thesis on the same topic.

MoRtHy - Monitor for Real-time Hyperproperties

This is a prototype implementation for sequential monitoring of HyperMTL.

The program will evaluate if a HyperMTL-B specification is satisfied by a list of traces. To run it, execute file morthy.py with two arguments: (1) input file of the HyperMTL specification formatted as described below, (2) input file with list of traces.

Example: ./morthy.py specs/spectest_a traces/traces_a.txt -v

Other command options: -v verbosity, -d <sampleperiod> enable discretization, -o <outputfile> set output for full result

Install Requirements

Written in python 3, the implementation builds on the Reelay package from Dogan Ulus.

Additional requirements are the python package sly and all inherited requirements from Reelay, namely a C++ compiler, the boost libraries for C++ and pybind11.

Specification Format

The specification format is based on the rye format for reelay.

Example: forall{1}forall{2}G[0:inf](F[0.5:1.5]{x_1} -> F[1:3]{x_2})

Use the following supported operators:

Trace quantifiers: forall{n} or exists{n}, with ascending trace variable numbers n

Temporal operators: eventually[a:b] , F[a:b] , once[a:b] , P[a:b] , always[a:b] , G[a:b] , historically[a:b] , H[a:b] , since[a:b] , S[a:b]

Logic operators: not , ! , and , && , or , || , implies, ->

Atomic propositions: {p_n}, n must be a valid trace variable number

Trace Format

Traces are implemented as lists of dicts, each dict represents a timestamp:

[
  {'time':0, 'x':True},
  {'time':0.6, 'x':False},
  {'time':1.5, 'x':True},
  {'time':2.2, 'x':False},
  {'time':2.67, 'x':True},
  {'time':5, 'x':False},
]

The input file for the trace set is then a list of traces.

Semantics

We assume interval semantics for the traces, meaning the value of x is constant until the next timestamp.

Acknowledgements

This implementation was built in the process of my thesis Monitoring Real-Time Hyperproperties.

Reelay was created by Dogan Ulus (Online monitoring of metric temporal logic using sequential networks).