Skip to content

This repository contains the benchmarks for neural network controller verification project using reach set computation.

Notifications You must be signed in to change notification settings

souradeep-111/Neural-Network-Controller-Verification-Benchmarks-HSCC-2019

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 

Repository files navigation

Author : Souradeep Dutta

Contact: [email protected]

This directory contains some sample benchmarks systems for the ARCH 2019 AI Model Verification category. The models used here were a part of the HSCC 2019 paper on 'Reachability Analysis for Neural Feedback Systems using Regressive Polynomial Rule Inference'

In total there are 11 benchmarks in the directory : ./Benchmarks/Ex_XXX/ The benchmarks are present as Matlab simulation scripts to produce trajectories starting from some initial set. To produce the trajectories run the script : 'simulate_with_NN.m' inside each folder. The details of the ODE model can be found inside : 'system_eq_dis.m'

FOR_ARCH_2019 Please consider benchmarks 7,9,10 .

For verification purpose, we wish to compute reach sets over some bounded horizon. The initial sets, and the bounded time horizon can be found in the files './Benchmarks/Ex_XXX/simulate_with_NN.m'

The description of the network format, in case it is required to be exported to some other format can be found in : https://github.com/souradeep-111/sherlock/blob/master/sherlock-network-format.pdf

Please feel free to reach out if you have further questions !

Cite the following paper if you use these benchmarks :

Reachability Analysis for Neural Feedback Systems using Regressive Polynomial Rule Inference, by Souradeep Dutta, Xin Chen and Sriram Sankaranarayanan HSCC 2019, Montreal, Canada

Bibtex:

@inproceedings{Dutta_Others__2019__Reachability, author = { Souradeep Dutta and Xin Chen and Sriram Sankaranarayanan }, title = { Reachability Analysis for Neural Feedback Systems using Regressive Polynomial Rule Inference }, booktitle = { Hybrid Systems: Computation and Control (HSCC) }, year = { 2019 }, pages = { TBA }, publisher = { ACM Press }}

About

This repository contains the benchmarks for neural network controller verification project using reach set computation.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published