https://github.com/panos108/gp-rto-ei
Real-time optimization using exploration strategies
https://github.com/panos108/gp-rto-ei
bayesian-optimization dynamic-optimization gaussian-processes real-time-operating-systems
Last synced: 7 months ago
JSON representation
Real-time optimization using exploration strategies
- Host: GitHub
- URL: https://github.com/panos108/gp-rto-ei
- Owner: panos108
- Created: 2020-06-19T14:35:56.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2021-02-19T16:45:20.000Z (over 5 years ago)
- Last Synced: 2025-03-30T17:11:15.273Z (about 1 year ago)
- Topics: bayesian-optimization, dynamic-optimization, gaussian-processes, real-time-operating-systems
- Language: Python
- Homepage:
- Size: 77.4 MB
- Stars: 5
- Watchers: 2
- Forks: 4
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# GP-RTO-EI
We present an RTO algorithm that relies on trust-region ideas in order to expedite and robustify convergence, and uses Bayesian optimization through Gaussian processes as a workhorse. We explore Expected Improvement and Upper Confidece Bound as adquisition functions, and adjust the size of the trust region based on the Gaussian processes’ ability to capture the plant-model mismatch in the cost and constraints. We draw parallels to expensive black-box optimization, hybrid modelling, reinforcement learning, and dual control which are all present in the proposed approach. Finally, we illustrate this new modifier-adaptation scheme on benchmark problems.