Reinforcement learning in extensive form games with incomplete information: the bargaining case study