Session Initiation Protocol (SIP) is considered as a signaling protocol for IP multimedia subsystem (IMS). IMS is introduced by 3rd generation partnership project (3GPP) as signaling foundation in next generation networks (NGN). Despite having the features such as: text based, IP based, independent of the data transmission, support for mobility and end-to-end, the SIP protocol has not suitable mechanism to deal with overload. Therefore, many mechanisms are proposed to control overload in SIP networks. One of their most famous is occupancy CPU (OCC) that is used in many researches. In traditional OCC, the value of parameters is indicated and they are used in subsequent documents. In this paper, optimal parameters value is obtained by Q-learning algorithm. Because modeling a large SIP network is impossible by mathematical relationships and it is a heuristic problem, Q-learning is the best method to compute the parameters. The simulation results demonstrate that the Q-learning output is comparable with traditional OCC.
