Reach-SDP: Reachability Analysis of Closed-Loop Systems with Neural Network Controllers via Semidefinite Programming

This benchmark suite presents a detailed description of a series of closed-loop control systems with artificial neural network controllers. In many applications, feed-forward neural networks are heavily involved in the implementation of controllers by learning and representing control laws through several methods such as model predictive control (MPC) and reinforcement learning (RL). The type of networks that we consider in this manuscript are feed-forward neural networks consisting of multiple hidden layers with ReLU activation functions and a linear activation function in the output layer. While neural network con- trollers have been able to achieve desirable performance in many contexts, they also present a unique challenge in that it is difficult to provide any guarantees about the correctness of their behavior or reason about the stability a system that employs their use. Thus, from a controls perspective, it is necessary to verify them in conjunction with their corresponding plants in closed-loop. While there have been a handful of works proposed towards the verification of closed-loop systems with feed-forward neural network controllers, this area still lacks attention and a unified set of benchmark examples on which verification techniques can be evaluated and compared. Thus, to this end, we present a range of closed-loop control systems ranging from two to six state variables, and a range of controllers with sizes in the range of eleven neurons to a few hundred neurons in more complex systems.

Download Full-text

Verisig 2.0: Verification of Neural Network Controllers Using Taylor Model Preconditioning

Computer Aided Verification - Lecture Notes in Computer Science ◽

10.1007/978-3-030-81685-8_11 ◽

2021 ◽

pp. 249-262

Author(s):

Radoslav Ivanov ◽

Taylor Carpenter ◽

James Weimer ◽

Rajeev Alur ◽

George Pappas ◽

...

Keyword(s):

Neural Network ◽

Closed Loop ◽

State Of The Art ◽

Taylor Model ◽

Shrink Wrapping ◽

Closed Loop Systems ◽

Verification Tool ◽

Extensive Evaluation ◽

Verification Tools ◽

Neural Network Controllers

AbstractThis paper presents Verisig 2.0, a verification tool for closed-loop systems with neural network (NN) controllers. We focus on NNs with tanh/sigmoid activations and develop a Taylor-model-based reachability algorithm through Taylor model preconditioning and shrink wrapping. Furthermore, we provide a parallelized implementation that allows Verisig 2.0 to efficiently handle larger NNs than existing tools can. We provide an extensive evaluation over 10 benchmarks and compare Verisig 2.0 against three state-of-the-art verification tools. We show that Verisig 2.0 is both more accurate and faster, achieving speed-ups of up to 21x and 268x against different tools, respectively.

Download Full-text

Neural Computation in Embodied Closed-Loop Systems for the Generation of Complex Behavior: From Biology to Technology

10.3389/978-2-88945-605-5 ◽

2018 ◽

Keyword(s):

Closed Loop ◽

Neural Computation ◽

Complex Behavior ◽

Closed Loop Systems

Download Full-text

METHOD OF SYNTHESIS OF CLOSED-LOOP SYSTEMS OF ACTIVE SHIELDING MAGNETIC FIELD OF POWER TRANSMISSION LINES

Tekhnichna Elektrodynamika ◽

10.15407/techned2016.04.008 ◽

2016 ◽

Vol 2016 (4) ◽

pp. 8-10 ◽

Cited By ~ 1

Author(s):

B.I. Kuznetsov ◽

◽

A.N. Turenko ◽

T.B. Nikitina ◽

A.V. Voloshko ◽

...

Keyword(s):

Magnetic Field ◽

Transmission Lines ◽

Closed Loop ◽

Power Transmission ◽

Power Transmission Lines ◽

Closed Loop Systems ◽

Active Shielding

Download Full-text

102-LB: Cross-Study Comparisons Done Right: An Illustration Using Two Pivotal Trials of Closed-Loop Systems

Diabetes ◽

10.2337/db20-102-lb ◽

2020 ◽

Vol 69 (Supplement 1) ◽

pp. 102-LB

Author(s):

MARC D. BRETON ◽

ROY BECK ◽

RICHARD M. BERGENSTAL ◽

BORIS KOVATCHEV

Keyword(s):

Closed Loop ◽

Closed Loop Systems

Download Full-text

Time in range for multiple technologies in type 1 diabetes: a systematic review and network meta-analysis

10.2337/figshare.12116691 ◽

2020 ◽

Author(s):

Anthony Pease ◽

Clement Lo ◽

Arul Earnest ◽

Velislava Kiriakova ◽

Danny Liew ◽

...

Keyword(s):

Systematic Review ◽

Type 1 Diabetes ◽

Closed Loop ◽

Meta Analysis ◽

Management Strategies ◽

Glucose Testing ◽

Closed Loop Systems ◽

Percent Time ◽

Time In Range

Background: Time-in-range is a key glycaemic metric, and comparisons of management technologies for this outcome are critical to guide device selection. Purpose: We conducted a systematic review and network meta-analysis to compare and rank technologies for time in glycaemic ranges. Data sources: We searched All Evidenced Based Medicine Reviews, CINAHL, EMBASE, MEDLINE, MEDLINE In-Process and other non-indexed citations, PROSPERO, PsycINFO, PubMed, and Web of Science until 24 April, 2019. Study selection: We included randomised controlled trials >2 weeks duration comparing technologies for management of type 1 diabetes in adults (>18 years of age), excluding pregnant women. Data extraction: Data were extracted using a predefined template. Outcomes were percent time with sensor glucose levels 3.9–10.0mmol/l (70–180mg/dL), >10.0mmol/L (180mg/dL), and <3.9mmol/L (70mg/dL). Data synthesis: We identified 16,772 publications, of which 14 eligible studies compared eight technologies comprising 1,043 participants. Closed loop systems lead to greater percent time-in-range than any other management strategy and was 17.85 (95% predictive interval [PrI] 7.56–28.14) higher than usual care of multiple daily injections with capillary glucose testing. Closed loop systems ranked best for percent time-in-range or above range utilising surface under the cumulative ranking curve (SUCRA–98.5 and 93.5 respectively). Closed loop systems also ranked highly for time below range (SUCRA–62.2). Limitations: Overall risk of bias ratings were moderate for all outcomes. Certainty of evidence was very low. Conclusions: In the first integrated comparison of multiple management strategies considering time-in-range, we found that the efficacy of closed loop systems appeared better than all other approaches.

Download Full-text