Learning for Safety-Critical Control with Control Barrier Functions.
Citations
1,977 citations
132 citations
105 citations
Cites methods from "Learning for Safety-Critical Contro..."
...Learning-based methods [6], [7] have been developed to address the conservatism of the robust strategies, but these can require extensive offline training to substantially improve the model....
[...]
62 citations
51 citations
References
2,391 citations
"Learning for Safety-Critical Contro..." refers background or methods in this paper
...In particular, this framework iteratively alternates between running experiments with an intermediate controller (or roll-outs in reinforcement learning (Kober et al. (2013))) to collect data, and using the newly collected data to synthesize a new controller....
[...]
...Learning-based approaches have already shown great promise for controlling systems with uncertain models (Schaal and Atkeson (2010); Kober et al....
[...]
...accurate model of the system, and thus model uncertainty can lead to degradation of these guarantees (Kolathaya and Ames (2018))....
[...]
...Learning-based approaches have already shown great promise for controlling systems with uncertain models (Schaal and Atkeson (2010); Kober et al. (2013); Khansari-Zadeh and Billard (2014); Cheng et al. (2019); Taylor et al. (2019b); Shi et al. (2019))....
[...]
2,186 citations
"Learning for Safety-Critical Contro..." refers background in this paper
...We look to achieve safety defined in terms of set invariance (Blanchini (1999); Ames et al. (2019)), which is an area of active research at the intersection of machine learning and control theory (Berkenkamp et al. (2016); Fisac et al. (2018))....
[...]
1,977 citations
"Learning for Safety-Critical Contro..." refers background in this paper
...Under this assumption, for any initial state x0 ∈ Rn, there exists a time interval of existence, I(x0) = [0, τmax), such that there is a unique solution, x : I(x0)→ Rn, satisfying (1) with x(0) = x0 (Perko (2013))....
[...]
1,931 citations
"Learning for Safety-Critical Contro..." refers methods in this paper
...To motivate our learning framework, consider a simple approach for learning a and b via supervised regression (Györfi et al. (2006)): an experiment is conducted using a nominal controller to collect data and learn functions that approximate a and b via supervised regression....
[...]
...To motivate our learning framework, consider a simple approach for learning a and b via supervised regression (Györfi et al. (2006)): an experiment is conducted using a nominal controller to collect data and learn functions that approximate a and b via supervised regression....
[...]
1,925 citations
"Learning for Safety-Critical Contro..." refers methods in this paper
...This assumption on the true system naturally follows if the true system retains the same actuation capability as the model, with more technical details provided in Sastry (1999)....
[...]