Automobiles increasingly include a multitude of replaceable and configurable hardware and software components. The pieces need to interoperate with each other in real-time, with their environment, and with consumer computing devices. Future cars may be reconfigured on a daily basis, according to e.g. car rental agreements.

Current real-time systems are not well suited to highly dynamic and even chaotic environments. Analyzing the problem to death and assigning a fixed schedule may be feasible for closed systems such as the brakes, but cannot be done for every possible addon combination and interaction with consumer devices.

This paper proposes a framework for adapting component interaction based on context histories and model based prediction. Tasks to be performed by a combination of software and hardware components are described in a pattern language, called the partiture. The partiture, analogous to the score of an orchestra, is used by the conductor to automatically configure, schedule, monitor, and adapt computation and communication of components.

Preliminary results demonstrate how a simple stochastic process can adapt a distributed entertainment scenario. This paper extends the concept to automotive applications with XML based interoperation, simple configuration, adaptive real-time, and security.