A naukri.com initiative
Arxiv
2d
43
Image Credit: Arxiv
Reachable Polyhedral Marching (RPM): An Exact Analysis Tool for Deep-Learned Control Systems
- Neural networks used in robotics require analysis of learned behaviors and their impact on closed-loop performance.
- Researchers have developed the Reachable Polyhedral Marching (RPM) algorithm for analyzing neural networks that implement piecewise-affine functions.
- The algorithm enables the computation of control invariant sets and regions of attraction (ROAs) for feedforward neural networks with ReLU activation.
- The approach showcases the ability to find non-convex control invariant sets and ROAs, as demonstrated in examples with learned oscillator and pendulum models.
Read Full Article
2 Likes
For uninterrupted reading, download the app