Learning-based testing (LBT) is a paradigm for fully automated requirements testing that combines machine learning with model-checking techniques. LBT has been shown to be effective for unit and integration testing of safety critical components in cyber-physical systems, e.g. automotive ECU software. We consider the challenges faced, and some initial results obtained in an effort to scale up LBT to testing co-operative open cyber-physical systems-of-systems (CO-CPS). For this we focus on a case study of testing safety and performance properties of multi-vehicle platoons.