Safety Considerations in Deep Control Policies with Safety Barrier Certificates Under Uncertainty

Hirshberg, Tom; Vemprala, Sai; Kapoor, Ashish

doi:10.1109/iros45743.2020.9341315

Cited by 3 publications

(4 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other recent works focused on control-oriented solutions to the safe learning problem based on the use of Barrier functions [21,33,60], robust MPC [63] or via projection on safe sets [29] are worth examining for interested readers.…”

Section: Control-based Solutions For Safe Learningmentioning

confidence: 99%

Learning against uncertainty in control engineering

Alamir

2022

Annual Reviews in Control

View full text Add to dashboard Cite

Section: Control-based Solutions For Safe Learningmentioning

confidence: 99%

Learning against uncertainty in control engineering

Alamir

2022

Annual Reviews in Control

View full text Add to dashboard Cite

“…x(t k ) = x(t k ), (16) where PC([t k , t k + T ], R m ) represents the set of all piece-wise continuous functions ϕ…”

Section: Learning-based Model Predictive Control With Cross-entropy M...mentioning

confidence: 99%

“…It essentially projects the desired trajectory into a set of safety-preserving controls whenever safety is threatened, so as to ensure safety and meanwhile keep the behavior as proactive as possible. Similarly, the control barrier function (CBF) is utilized to design the safety barrier certificates in [15] [16] to minimally modify an existing controller to formally satisfy collision avoidance constraints via a constrained quadratic program (QP). While robust control under bounded uncertainties achieves safety and meanwhile preserves the control performance of the existing controller, the adaptation to disturbances is not considered in [15] [16].…”

Section: Introductionmentioning

confidence: 99%

“…Similarly, the control barrier function (CBF) is utilized to design the safety barrier certificates in [15] [16] to minimally modify an existing controller to formally satisfy collision avoidance constraints via a constrained quadratic program (QP). While robust control under bounded uncertainties achieves safety and meanwhile preserves the control performance of the existing controller, the adaptation to disturbances is not considered in [15] [16]. The CBF and control Lyapunov function (CLF) are integrated into a QP to achieve both safety and tracking stability in [17], but without a minimal intervention scheme.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Safe Learning-based Gradient-free Model Predictive Control Based on Cross-entropy Method

Zheng,

Yang,

et al. 2021

Preprint

View full text Add to dashboard Cite

In this paper, a safe and learning-based control framework for model predictive control (MPC) is proposed to optimize nonlinear systems with a gradient-free objective function under uncertain environmental disturbances. The control framework integrates a learning-based MPC with an auxiliary controller in a way of minimal intervention. The learning-based MPC augments the prior nominal model with incremental Gaussian Processes to learn the uncertain disturbances. The cross-entropy method (CEM) is utilized as the sampling-based optimizer for the MPC with a gradient-free objective function. A minimal intervention controller is devised with a control Lyapunov function and a control barrier function to guide the sampling process and endow the system with high probabilistic safety. The proposed algorithm shows a safe and adaptive control performance on a simulated quadrotor in the tasks of trajectory tracking and obstacle avoidance under uncertain wind disturbances.

show abstract