Dynamic customization of time-sensitive networking (TSN) with digital twins

A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations comprising:

• accessing at a Time-Aware Scheduler (TAS) of a Time-Sensitive Networking (TSN) network a plurality of data packets, each data packet including a data packet priority that defines an order for transmitting the plurality of data packets by the TSN network;
• grouping the plurality of data packets into a cluster according to the data packet priorities, the grouping being performed by a digital twin that represents the TAS and that continuously monitors network traffic to recognize a current TSN network operation state, the cluster identifying a first TSN network state;
• applying a first machine-learning model to the cluster to determine a TAS cycle policy that defines a timing schedule for data packet transmission in the TSN network, wherein the first machine-learning model is selected, by the digital twin, from a plurality of reinforcement learning (RL) models that are respectively trained for different TSN network operation states including a default network state, a congestion network state, and a new priority network state;
• and
• providing the TAS cycle policy to components of the TSN network, including transmitting a configuration message from the digital twin to a physical TAS to dynamically reconfigure time slots of a TAS cycle based on the recognized TSN network operation state.