Ambient-Oriented Programming

In the general research area of pervasive computing and ambient intelligence, our research tends to focus on so-called mobile networks. Mobile networks fundamentally differ from traditional, stationary networks in the following two ways:

1. Whereas traditional distributed systems are often composed of heavyweight, computationally intensive nodes, mobile networks tend to consist of small, computationally scarce and mobile devices. A device may be mobile simply because it is small (e.g. a cell phone) or because it is embedded in another mobile devices (e.g. a car's on-board computer).
2. In traditional distributed systems, components are interconnected via fast, wired, reliable communication links. In a mobile network, connections are usually wireless, slower and less reliable. Moreover, the biggest consequence of the wireless connections is that the communication range of devices becomes limited.

Given these unavoidable, fundamental properties of mobile network, our research is founded on the observation of a number of so-called hardware phenomena: phenomena which have a profound impact on software, and which cannot be hidden from either developer or end-user because they are innate to the underlying hardware. The hardware phenomena we observe are:

• Connection Volatility: Two processes that perform a meaningful task together on two cooperating devices cannot assume a stable connection. The limited communication range of the wireless technology combined with the fact that users can move out of range can result in broken connections. However, upon re-establishing a broken connection, users typically expect the task to resume. In other words, they expect the task to be performed in the presence of a volatile connection.
• Ambient Resources: If a user moves with his mobile device, remote resources become dynamically (un)available in the environment because the availability of a resource may depend on the location of the device. This is in contrast with stationary networks in which references to remote resources are obtained based on the explicit knowledge of the availability of the resource. In the context of mobile networks, the resources are said to be ambient.
• Autonomy: Most distributed applications today are developed using the client-server approach. The server often plays the role of a “higher authority” which coordinates interactions between the clients. In mobile networks, and especially in mobile ad hoc networks, a connection to such a “higher authority” is not always possible. Every device acts as an autonomous computing unit.
• Natural Concurrency: In theory, distribution and concurrency are two different phenomena. For instance in a client-server setup, a client might wait for the results of a request to the server in order to resume its computation. Hence, in theory a distributed system is not necessarily a concurrent one. However, even in the extreme case where both communicating devices run a single threaded program, their autonomy implies that the resulting task is a concurrent one. Moreover, the trend of software getting ever more multi-threaded will also manifest itself on mobile devices. As a result, concurrency is a natural phenomenon in software running on mobile networks.