This paper deals with a real-time scheduling method for holonic manufacturing systems (HMS). In the previous paper, a real-time scheduling method based on utility values has been proposed and applied to the HMS. In the proposed method, all the job holons and the resource holons firstly evaluate the utility values for the cases where the holon selects the individual candidate holons for the next machining operations. The coordination holon secondly determine a suitable combination of the resource holons and the job holons which carry out the next machining operations, based on the utility values. Multi-agent reinforcement learning is newly proposed and implemented to the job holons and the resource holons, in order to improve their capabilities for evaluating the utility values of the candidate holons. The individual job holons and resource holons evaluate the suitable utility values according to the status of the HMS, by applying the proposed learning method.