In this article, we present a framework ABOS based on acollaborative multi-agent system (MAS) for multi-skill health care scheduling using metaheuristics. In this framework, each agent preforms his actions autonomously in the search space of a scheduling optimization problem. Information about patient scheduling is shared between agents who collaborate through the dynamic environment. The objective is to allow the agents to adapt their decisions using the Reinforcement Learning approach according to the acquired experience with the interaction with the other agents and the environment. The aim of this interaction between agents is to enhance the quality of the solutions provided by the agents from the search space. Experiments were performed using real data provided by the adult emergency department (AED) of Lille university hospital center (LUHC). The simulation results confirm that the integration of Machine Learning in agent behaviors impacts the quality of the scheduling solution. Collaboration between agents in ”friend” or ”enemy” mode influences the quality of the solution as well and thus impacts the health care patient pathway.