Reachable Set Estimation and Verification for Neural Network Models of Nonlinear Dynamic Systems

Xiang, Weiming; Lopez, Diego Manzanas; Musau, Patrick; Johnson, Taylor T.

doi:10.1007/978-3-319-97301-2_7

Cited by 24 publications

(17 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An important factor making our approach potentially faster and more scalable than the Verisig and SMC-based approaches is, our approach can efficiently compute the exact reachable set of DNNs on multi-core platforms. Therefore, our 105:19 [11,19,27,28,34,36]. The early polyhedron-based approach has been extended for safety verification of neural network controlled systems in [36] in which the plant is assumed to be linear and discrete.…”

Section: Adaptive Cruise Control Systemmentioning

confidence: 99%

Safety Verification of Cyber-Physical Systems with Reinforcement Learning Control

Tran

Cai

Diego

et al. 2019

ACM Trans. Embed. Comput. Syst.

Self Cite

View full text Add to dashboard Cite

This paper proposes a new forward reachability analysis approach to verify safety of cyber-physical systems (CPS) with reinforcement learning controllers. The foundation of our approach lies on two efficient, exact and over-approximate reachability algorithms for neural network control systems using star sets, which is an efficient representation of polyhedra. Using these algorithms, we determine the initial conditions for which a safety-critical system with a neural network controller is safe by incrementally searching a critical initial condition where the safety of the system cannot be established. Our approach produces tight over-approximation error and it is computationally efficient, which allows the application to practical CPS with learning enable components (LECs). We implement our approach in NNV, a recent verification tool for neural networks and neural network control systems, and evaluate its advantages and applicability by verifying safety of a practical Advanced Emergency Braking System (AEBS) with a reinforcement learning (RL) controller trained using the deep deterministic policy gradient (DDPG) method. The experimental results show that our new reachability algorithms are much less conservative than existing polyhedra-based approaches. We successfully determine the entire region of the initial conditions of the AEBS with the RL controller such that the safety of the system is guaranteed, while a polyhedra-based approach cannot prove the safety properties of the system.

show abstract

Section: Adaptive Cruise Control Systemmentioning

confidence: 99%

Safety Verification of Cyber-Physical Systems with Reinforcement Learning Control

Tran

Cai

Diego

et al. 2019

ACM Trans. Embed. Comput. Syst.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The result of cross multiplication Feedforward neural networks or Multilayer Perceptrons (MLPs) [15] are the base of the Deep Learning Model. The main objective of the feed-forward network is to define the mapping of input , − ( ; ) categories and to estimate the value of the parameter θ, which is the result of the best function estimate [16][17].…”

Section: Neural Networkmentioning

confidence: 99%

Spatial Based Deep Learning Autonomous Wheel Robot Using CNN

Prasetyo

Nambo

Prasetya³

et al. 2020

LKJITI

View full text Add to dashboard Cite

The development of technology is growing rapidly; one of the most popular among the scientist is robotics technology. Recently, the robot was created to resemble the function of the human brain. Robots can make decisions without being helped by humans, known as AI (Artificial Intelligent). Now, this technology is being developed so that it can be used in wheeled vehicles, where these vehicles can run without any obstacles. Furthermore, of research, Nvidia introduced an autonomous vehicle named Nvidia Dave-2, which became popular. It showed an accuracy rate of 90%. The CNN (Convolutional Neural Network) method is used in the track recognition process with input in the form of a trajectory that has been taken from several angles. The data is trained using Jupiter's notebook, and then the training results can be used to automate the movement of the robot on the track where the data has been retrieved. The results obtained are then used by the robot to determine the path it will take. Many images that are taken as data, precise the results will be, but the time to train the image data will also be longer. From the data that has been obtained, the highest train loss on the first epoch is 1.829455, and the highest test loss on the third epoch is 30.90127. This indicates better steering control, which means better stability.

show abstract

“…To provide strong safety and reliability guarantees, several works in the literature focus on applying formal verification techniques (e.g., model checking) to verify pre-trained MLbased controllers' formal safety properties. Representative examples of this approach are the use of SMT-like solvers [13], [14], [15] and hybrid-system verification [16], [17], [18]. However, these techniques only assess a given ML-based controller's safety rather than design or train a safe agent.…”

Section: Introductionmentioning

confidence: 99%

Provably Correct Training of Neural Network Controllers Using Reachability Analysis

Sun,

Shoukry

2021

Preprint

View full text Add to dashboard Cite

In this paper, we consider the problem of training neural network (NN) controllers for cyber-physical systems (CPS) that are guaranteed to satisfy safety and liveness properties. Our approach is to combine model-based design methodologies for dynamical systems with data-driven approaches to achieve this target. Given a mathematical model of the dynamical system, we compute a finite-state abstract model that captures the closedloop behavior under all possible neural network controllers. Using this finite-state abstract model, our framework identifies the subset of NN weights that are guaranteed to satisfy the safety requirements. During training, we augment the learning algorithm with a NN weight projection operator that enforces the resulting NN to be provably safe. To account for the liveness properties, the proposed framework uses the finitestate abstract model to identify candidate NN weights that may satisfy the liveness properties. Using such candidate NN weights, the proposed framework biases the NN training to achieve the liveness specification. Achieving the guarantees above, can not be ensured without correctness guarantees on the NN architecture, which controls the NN's expressiveness. Therefore, and as a corner step in the proposed framework is the ability to select provably correct NN architectures automatically.

show abstract

Reachable Set Estimation and Verification for Neural Network Models of Nonlinear Dynamic Systems

Cited by 24 publications

References 47 publications

Safety Verification of Cyber-Physical Systems with Reinforcement Learning Control

Safety Verification of Cyber-Physical Systems with Reinforcement Learning Control

Spatial Based Deep Learning Autonomous Wheel Robot Using CNN

Provably Correct Training of Neural Network Controllers Using Reachability Analysis

Contact Info

Product

Resources

About