About the project

Background

RoboTest is a five-year EPSRC-funded project involving researchers from The University of York and The University of Sheffield, developing a model-based framework for sound and automated testing of robotic systems.

Currently, robotics is at the heart of the UK economic plans. A limiting factor, however, is safety, which is a challenge for design and assurance.

Current practice for robot controller development is typically based on standard state machines, without formal semantics. For design, the state machine guides the development of simulation and ad hoc simulation experiments. For implementation, adjustments are normally required to cater for the reality gap between the simulation and the real world.

Here, another set of running conditions may be considered for testing. There is no rigorous approach for the use of simulation and testing with impact on cost, maintainability, and reliability.

Aims

The aim is to tackle testing based on simulations, and the conversion of simulation tests for use on robots operating in the real world. Our models will be based on a domain-specific notation for mobile and autonomous robots interacting with a complex environment.

This notation is being developed in a separate, but related, project: RoboCalc. The notation supports specification of functional, timed, and probabilistic properties.

Our goal is to develop novel and sound test generation and execution techniques to support systematic and automated simulation experiments.

Soundness will provide a precise and reliable argument of what the experiments achieve, with respect, for example, to classes of faults. In this way, it will be more likely that we can find design flaws and faults more cheaply and early.

Impact

Since simulation is widely used in robotic design, improvements to productivity, effectiveness, and cost of running simulations will have a significant impact. We will consider several industrial simulation tools for robotics.

In addition, we will automatically convert simulation to deployment tests for use with the robots to increase traceability and reusability.

Within deployment tests, we intend also to explore the reality gap between the simulation and the real world. The purpose is to enrich current practice with a facility for automatic generation of sound simulations and tests.

Our framework will enable a transformative change by providing industry and academia with invaluable tools:

• Cost-effective techniques for simulation and testing.
• A pathway for regulators to consider testing requirements in safety arguments.
• Guidance for simulation-tool developers to improve their tools.