2020
DOI: 10.48550/arxiv.2012.14964
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

Control Barriers in Bayesian Learning of System Dynamics

Vikas Dhiman,
Mohammad Javad Khojasteh,
Massimo Franceschetti
et al.

Abstract: This paper focuses on learning a model of system dynamics online while satisfying safety constraints. Our objective is to avoid offline system identification or hand-specified models and allow a system to safely and autonomously estimate and adapt its own model during operation. Given streaming observations of the system state, we use Bayesian learning to obtain a distribution over the system dynamics. Specifically, we use a matrix variate Gaussian process (MVGP) regression approach with efficient covariance f… Show more

Help me understand this report
View published versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
8
0

Year Published

2021
2021
2021
2021

Publication Types

Select...
3

Relationship

0
3

Authors

Journals

citations
Cited by 3 publications
(8 citation statements)
references
References 50 publications
0
8
0
Order By: Relevance
“…Thus, "energy deficit" CBF defined by (11) has relative degree 2, from the voltage input, according to the perspective in [42], which inherits the backstepping change of variable ( 13) from [36].…”
Section: Nonovershooting Regulation By Backstepping For Multiple Cbfsmentioning
confidence: 99%
See 1 more Smart Citation
“…Thus, "energy deficit" CBF defined by (11) has relative degree 2, from the voltage input, according to the perspective in [42], which inherits the backstepping change of variable ( 13) from [36].…”
Section: Nonovershooting Regulation By Backstepping For Multiple Cbfsmentioning
confidence: 99%
“…Due to the addition of the solid phase temperature dynamics, the energy deficit σ in (11) is reformulated. We consider the following CBFs for the two-phase problem:…”
Section: B Nonovershooting Regulation and Guaranteed Safetymentioning
confidence: 99%
“…moving obstacles), which cannot be controlled. 3 The size of σ(X t ) is determined from the uncertainties in the disturbance, unmodeled dynamics [8], [10], [11], [20], and the prediction errors of the environmental variables [13], [21]. Examples of these cases include when the unmodeled dynamics are captured using statistical models such as Gaussian Processes 4 and when the motion of the environment variables are estimated using physics-based models such as Kalman filters [13], [21].…”
Section: A System Description and Design Specificationsmentioning
confidence: 99%
“…The mapping K is either (19) or (20), depending on the choice of control policies. In the next section, we present out main theorems.…”
Section: Safe Probabilitymentioning
confidence: 99%
See 1 more Smart Citation