SIP registration optimization in mobile environments using extended Kalman filter

The Internet is becoming the dominating medium for modern communications. The SIP protocol delivers the majority of VoIP services. The SIP usage in mobile environments becomes reality due to the widespread of portable devices together with dynamic lifestyle. Unfortunately, SIP is not designed to work in mobile environments where access network and IP addresses change frequently. Intermediate NAT devices change additionally the network parameters without notifying the end hosts. To eliminate disconnection due to change of the IP or port parameter, the SIP client must proactively register and update its current parameters in this way. The frequency of SIP registration is decisive for the potential disconnection. The current registrations are executed in constant intervals set regardless of the mobility and network properties. The result in mobile environments is wasted resources and overloaded links. A new method for optimization of the registration intervals is presented in this paper. The key idea is to set the registration intervals proportional to the probability for disconnection. The probability density function is built from the past disconnections. The new approach is based on extended Kalman filter. A simulation shows the qualities of the new method, where clear outperformance is demonstrated. The algorithm can be implemented in protocols like IKE, routing protocols, Mobile IP etc.