Theme 3: Services, Applications and Cloud Interactions
The TerraSwarm vision is one of composable services that can be dynamically recruited by applications. Formally, applications are dened as dynamic, distributed graphs of connected services. Both "dynamic" and "distributed" are important here; applications persist even as the individual components that comprise these applications change. This view elevates the concept of an integrated modular architecture (IMA), today's target for systems-of-systems design, from the system level to the enterprise level, and augments it with discovery and run-time adaptation.
Control as a Service. From the user perspective, the TerraSwarm provides (contextual) total awareness, which is enabled by a dynamically changing mixture of local and remote swarm sensors. Adaptive services will exploit these devices to improve accuracy and quality for the user. Ensuring that such adaptive services remain effective, efficient, and safe under dynamic restructuring is a challenging control problem. The TerraSwarm vision is to decentralize the design of such systems, improving their robustness and making them more adaptable and opportunistic. Control strategies will be synthesized on the y from goal specifications and constraints, a vision we call control as a service.
The Cloud as a Companion. A central challenge to be overcome is the imbalance between the massive amounts of information that could be collected and the time-sensitive interests and needs of the user(s). A naïve approach is to collect and store all data, and have cloud-based services distill the information for user consumption. But the most interesting services will need the right (contextual) data at the right time and the right place. Closed-loop cyber-physical interactions will not tolerate the latencies incurred by cloudbased archiving and indexing. Moreover, the vast data flood that will emerge from the Terraswarm make this naïve approach far too costly, even with huge advances in storage technology.
In the TerraSwarm environment, resources are recruited opportunistically based on availability and need, with the objective of providing the best possible experience to the user. Data produced locally will be maximally leveraged locally. Nevertheless, the cloud plays an essential role. A key goal of this theme is to "wake up" the cloud, giving it a physical rather than just cognitive presence; rather than just providing information, a TerraSwarm system will affect our physical environment.
As with social networks and information search technologies, the cloud participates by aggregating data from a multiplicity of sources, something not possible on a single physical device, no matter how much computation and memory capability it has. The cloud is not just a computation and memory resource; it is an information aggregator and a service synthesizer. Data aggregation allows us to shift feedback control from the system level to the enterprise level.
Structured Data Summarization. The vast quantity and variety of swarm data will require new approaches for correlating, interpreting, and displaying data in a meaningful way. We plan to develop effective mechanisms for managing swarm data.
Secure and Safe Swarms. The web and social media have opened the floodgates of personal information available about us even to strangers. Even as our culture is only starting to learn to deal with the consequences of that information flood, that flood is about to be itself overwhelmed by data streams from physical sensors. The TerraSwarm vision is that security and privacy must be built into the very core of service definition.
The TerraSwarm project will use a system theoretic formulation to address privacy concerns, defining filters that release useful information without compromising privacy. Our proposed approach relies on the notion of differential privacy, which provides strong privacy guarantees against adversaries with arbitrary side information.
We will also examine potential data leakage introduced by composable services through side channels such as timing and power consumption. Fortunately, there are synergies. For example, temporal isolation may be introduced to guarantee resources to safety-critical services, but it can also be used to prevent side-channel attacks, where private information is deduced from temporal variations in software execution.
We will explore the use of security-related technologies and techniques such as static analysis, hazard analysis, and elliptic curve cryptography to implement effective security approaches. We will leverage existing research in the area of distributed storage to inform the design of cloud-based swarm applications that need strong guarantees of security despite their reliance on physically insecure infrastructure.
Services, Applications and Cloud Interactions
- Richard Murray (Leader)
- Jeffrey A. Bilmes (Co-Leader)
- See the Services Overview for a list of other faculty.
You are not logged in or do not have access to most of this workgroup. Admins, to edit this page, see the profile.