The main focus of this paper is to study and develop an efficient learning policy to address the exploration vs. exploitation dilemma in a multi-objective foraging task in swarm robotics domain. An efficient learning policy called FIFO-list is proposed to tackle the above mentioned problem. The proposed FIFO-list is a model-based learning policy which can attain nearoptimal solutions. In FIFO-list, the swarm robots maintains a dynamic list of recently visited states. States that are included in the list are banned from exploration by the swarm robots regardless of the Q(s, a) values associated with those states. The FIFO list is updated based on First-In-First-Out (FIFO) rule, meaning the states that enters the list first will exit the list first. The recently visited states will remain in the list for a dynamic number of time-steps which is determined by the desirability of the visited states. © 2013 IEEE.