Abstract:In order to formally verify robotic controllers, we must tackle the inherent uncertainty of sensing and actuation in a physical environment. We can model uncertainty using stochastic hybrid systems, which combine discrete jumps with continuous, stochastic behaviour. The verification of these systems is intractable for state-exploration based approaches, so we instead propose a deductive verification approach. As a first step towards a deductive verification tool, we present a mechanisation of Brownian motion within Isabelle/HOL. For this, we mechanise stochastic kernels and Markov semigroups, which allow us to specify a range of processes with stationary, independent increments. Further, we prove the Kolmogorov-Chentsov theorem, which allows us to construct Hölder continuous modifications of processes that satisfy certain bounds on their expectation. This paves the way for modelling and verifying stochastic hybrid systems in Isabelle/HOL.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: **How to formally verify robot controllers to deal with the inherent uncertainties in sensing and execution in the physical environment**.
Specifically, the author points out that the robot system operates in continuous time and space, and the verification of its control software must consider this uncertainty. To model this uncertainty, Stochastic Hybrid Systems (SHS) can be used, which combines discrete jumps with continuous stochastic behavior. However, traditional state - exploration - based verification methods are not feasible for SHS because the state space of these systems is too large, resulting in a complex and inaccurate verification process.
To solve these problems, the author proposes a deductive verification method instead of relying on automated, state - exploration - based methods. This method allows symbolic reasoning, thus avoiding the need for discretization, and guides the verification process through human input. In addition, deductive verification does not require approximate calculations and is not limited to linear Stochastic Differential Equations (SDEs), so it has broader applicability and stronger verification guarantees.
As the first step towards a deductive verification tool, the author has implemented the mechanization of Brownian motion in Isabelle/HOL. Specific contributions include:
1. **Constructing stochastic processes**: Stochastic processes are constructed through families of stochastic kernels, which generalize Markov chains.
2. **Formalizing the Kolmogorov - Chentsov theorem**: This theorem enables the construction of processes with almost everywhere continuous paths.
3. **Implementing Brownian motion**: Using the above results, Brownian motion is constructed in Isabelle/HOL.
These achievements lay the foundation for modeling and verifying stochastic hybrid systems in Isabelle/HOL. In this way, the author hopes to better handle the uncertainty problem in the physical environment in formal verification.
### Formula presentation
- **Definition of stochastic kernel**:
\[
\kappa: \Omega\times\mathcal{A}'\to\mathbb{R}^+
\]
Satisfying the following conditions:
- For all \(A'\in\mathcal{A}'\), \(\kappa(-, A')\) is \(\mathcal{A}\)-Borel measurable.
- For all \(\omega\in\Omega\), \(\kappa(\omega, -)\) is a measure on \((\Omega', \mathcal{A}')\).
- **Kolmogorov - Chentsov theorem**:
If a stochastic process \(X\) satisfies:
\[
E[d(X_t, X_s)^\alpha]\leq C|t - s|^{1+\beta}, \quad s, t\in\mathbb{R}^+
\]
Then there exists a modified version \(Y\) of \(X\) such that for any \(\gamma\in(0, \frac{\beta}{\alpha})\), the paths of \(Y\) are H\"older - \(\gamma\) continuous.
Through these formulas and theories, the author has successfully implemented Brownian motion and its related properties in Isabelle/HOL, providing a solid foundation for subsequent verification of stochastic hybrid systems.