Formal Verification of ROS-Based Robotic Applications Using Timed-Automata

Abstract

Robotic technologies are continuously transforming the domestic and the industrial environments. Recently the Robotic Operating System (ROS), an open source middleware framework, has been widely adopted both by industry and academia, becoming the de facto standard for developing robot applications. Guaranteeing the correct behaviour of robotic systems is, however, challenging due to their potential for parameterization and heterogeneity. Although different approaches exist, focusing on concrete domain spaces for specific scenarios, no general approach to reason about ROS systems has yet arisen. This paper proposes an approach to model and verify ROS systems using real time properties, focusing on one of the main features of ROS, the communication between nodes. It takes low-level parameters into account, such as queue sizes and timeouts, and uses timed automata as the modelling language. The physical robot Kobuki is used as a complex case study, over which properties are automatically verified using the UPPAAL model checker, enabling the identification of problematic parameter combinations.

Statistical Model Checking of Complex Robotic Systems

“…That is, robotic applications, initially specified in a robotic framework, need to be modeled first in RoboChart, then translated into CSP. An attempt to formalize ROS components is developed in [13] where UPPAAL is used to verify buffer-related properties (no overflow). Only the message passing part (publisher/subscriber) is modeled, manually, and crucial bounded response properties (e.g.…”

Formal Specification and Verification of Autonomous Robotic Systems

“…It is often assumed that a middleware is sound, and this is key to trusting the robotic system [75]. However, given the heterogeneity of the systems that can be produced using such architectures and their parametrisable nature, guaranteeing correct robot behaviour is challenging [86]. Thus, we have identified the following categories of nonand semi-formal methods.…”

“…However, they are time-consuming, and examine part of the program's state space, so they cannot be used to reason reliably about properties of the whole program. Moreover, field tests are potentially dangerous to life and the robot hardware Examples include the following uses of field tests [110,75,86,128] and/or simulations [110,9].…”

“…The existing tool-chain for G en oM has been extended to generate Fiacre models of a system, which can then be model-checked against LTL properties [75]. Another approach models ROS nodes, and the communication between them, as TA; which are checkd against TCTL properties in Uppaal [86]. They verify the lack of message buffer overflows for a selection of queue sizes in the open-source robotic application Kobuki.…”

Recent Trends in Formal Validation and Verification of Autonomous Robots Software

“…Despite its success, ROS provides little support when it comes to ease V&V of the software with formal models. There are some efforts to model its communication layer [Halder et al, 2017], or to verify some simple properties [Come et al, 2018;Meng et al, 2015;Wong and Kress-Gazit, 2017], but overall, the lack of structure required to write ROS nodes makes it rather difficult to extract anything worth verifying. [Bardaro et al, 2018] proposes to model the robot software in AADL and then synthesize code in ROS.…”