1: \begin{abstract}

2: The region of attraction characterizes well-behaved and safe operation of a nonlinear system and is hence sought after for verification.

3: %For nonlinear systems, estimating the region of attraction is part of a quantitative analysis for safety verification. While analytical methods scale badly for large problems, sampling-based methods are applicable to complex systems even with partially unknown dynamics. 

4: %Previous results, however, either failed to represent the region of attraction accurately, lacked probabilistic guarantees, or the necessary number of samples too grew with the complexity.

5: In this paper, a framework for probabilistic region of attraction estimation is developed that combines scenario optimization and converse theorems. With this approach, the probability of an unstable condition being included in the estimate is independent of the system's complexity, while convergence in probability to the true region of attraction is proven.

6: Numerical examples demonstrate the effectiveness for optimization-based control applications. Combining systems theory and sampling, the complexity of Monte--Carlo-based verification techniques can be reduced.

7: The results can be extended to arbitrary level sets of which the defining function can be sampled, such as finite-horizon viability.

8: Thus, the proposed approach is applicable and/or adaptable to verification of a wide range of safety-related properties for nonlinear systems including feedback laws based on optimization or learning. 

9: \end{abstract}

10: