The communication services are usually transmitted across heterogeneous networks with different access technologies, which makes QoS assurance for end-to-end deterministic communication services more complexed. Machine learning (ML)-based joint resource scheduling could reduce the algorithm complexity due to its capability to make optimized online scheduling and resource allocation among multiple networks based on the services and networks states. Therefore, this Recommendation specifies the QoS requirements, framework and operational procedures to support deterministic communication services across heterogeneous networks including IMT-2020 and beyond enabled by ML-based joint resource scheduling.