Mathematics > Optimization and Control
[Submitted on 28 Sep 2019 (this version), latest version 16 Mar 2020 (v2)]
Title:Guaranteed-Safe Approximate Reachability via State Dependency-Based Decomposition
View PDFAbstract:Hamilton Jacobi (HJ) Reachability is a formal verification tool widely used in robotic safety analysis for controlled nonlinear systems experiencing adversarial disturbances. Given a target set that represents a set of unsafe states, the Backward Reachable Tube (BRT) specifies the states from which a system is guaranteed to enter the target under the worst-case disturbance; therefore, the complement of the BRT guarantees safety. However, computing BRTs requires state space discretization, which suffers from exponential computational time and space complexity with respect to the state space dimension. Previously, system decomposition and projection techniques have been investigated for tackling this curse of dimensionality, but the trade off between applicability to a wider class of dynamics and degree of conservatism has been challenging. In this paper, we first propose a State Dependency Graph to represent the system dynamics, and then decompose the full system in a way that only dependent states are included in each subsystem, and "missing" states are treated as bounded disturbance. Thus for a large variety of dynamics in robotics, BRTs can be quickly approximated in lower-dimensional chained subsystems, while conservatism is preserved in the right direction to guarantee safety. We demonstrate our method with numerical experiments on the 4D Quadruple Integrator, and the 6D Bicycle, an important car model that has been intractable to analyze to the best of our knowledge.
Submission history
From: Anjian Li [view email][v1] Sat, 28 Sep 2019 07:06:11 UTC (4,388 KB)
[v2] Mon, 16 Mar 2020 22:49:59 UTC (4,433 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.