More and more people are watching media content over the internet or from own personal media servers, and this way is slowly taking over for traditional television broadcasts. With this, the desire and possibility to watch media content from any location at any time arises as people are often on the move. With todays technology everything from computers to television sets and even hand-held devices have access to the internet and capability for decoding and rendering video. The problem addressed in this master thesis arises when the user decides from one location to another location, for example from the living room to the bedroom. The user might want to continue to watch what he was already watching in the living room when he gets to the bedroom.
This problem gives root to another desired service; being able to move a media streaming session from one device, for example a television set, to another, for example a hand held PDA. This is often known as client side session migration or session hand-off.
Ideally, such functionality should work with as little user interaction and as seamlessly as possible. This thesis proposes a way of adding session migration functionality to an existing media streaming application, the so-called Personal Media Service (PMS). The PMS application uses the QuA middleware platform to automatically adapt media quality to the client's context. It is proposed that each device can be represented by a QuA service mirror. In this thesis the QuA functionality is expanded to also use the user's and devices' physical locations to decide what device the media server is to stream to.
As is shown through the design, implementation and testing done in this thesis, using planning based middleware for session migration by representing each device as a service mirror will work. Bandwidth, and user and device positions, allow the planning-based middleware to automatically determine which device will give the best user experience at any time. The tests show that performance of session migration using the QuA middleware does not scale very well for many available devices, but works su ciently well for few devices.