Neural barrier functions

Neural networks as barrier functions for stochastic discrete-time systems trained and verified using bound propagation.

To train:

python experiments/main.py --device=<cpu|cuda> --config-path=<config-path> --save-path=models/<model-name>.{state}.pth --task=train

To certify:

python experiments/main.py --device=<cpu|cuda> --config-path=<config-path> --save-path=models/<model-name>.{state}.pth --task=test

To plot:

python experiments/main.py --device=<cpu|cuda> --config-path=<config-path> --save-path=models/<model-name>.{state}.pth --task=plot

Authors

• Frederik Baymler Mathiesen - PhD student @ TU Delft

Funding and support

• TU Delft

Copyright notice:

Technische Universiteit Delft hereby disclaims all copyright interest in the program “neural-barrier-functions” (neural networks as barrier functions with bound propagation) written by the Frederik Baymler Mathiesen. Theun Baller, Dean of Mechanical, Maritime and Materials Engineering

© 2022, Frederik Baymler Mathiesen, HERALD Lab, TU Delft