Using CP in Automatic Test Generation for ABB Robotics' Paint Control System

Designing industrial robot systems for welding, painting, and assembly, is challenging because they are required to perform with high precision, speed, and endurance. ABB Robotics has specialized in building highly reliable and safe robotized paint systems based on an integrated process control system. However, current validation practices are primarily limited to manually designed test scenarios. A tricky part of this validation concerns testing the timing aspects of the control system, which is particularly challenging for paint robots that need to coordinate paint activation with the robot motion control. To overcome these challenges, we have developed and deployed a costeffective, automated test generation technique based on Constraint Programming, aimed at validating the timing behavior of the process control system. We designed a constraint optimization model in SICStus Prolog, using arithmetic and logic constraints including use of global constraints. This model has been integrated into a fully automated continuous integration environment, allowing the model to be solved on demand prior to test execution, which allows us to obtain the most optimal and diverse set of test scenarios for the present system configuration. After three months of daily operational use of the constraint model in our testing process, we have collected data on its performance and bug finding capabilities. We report on these aspects, along with our experiences and the improvements gained by the new testing process.