Improving the Usefulness of Context Information for IoT Applications: A Middleware-based Approach

Smarter pervasive and ubiquitous applications are on the rise. They leverage contextual information to make actions more relevant and dependable to the users and the Internet of Things (IoT) ecosystem. The adoption of context is making large leaps in time- and safety-critical applications like autonomous driving. While the sources of context information are also rapidly increasing to support this trend, they fail to provide consistent quality, performance and vary largely in features, e.g., presentation and interfaces. In large IoT ecosystem, untangling these complexities and responding to each context request with useful context information is expensive and not scalable. We investigate and propose a novel mechanism to provide Context Management Systems (CMS) with (a) adaptive context caching for reusing and repurposing context information for fast concurrent access, and (b) cost- and quality-aware context provider selection to supplement caching with means to maximize effectiveness. Unlike any previous work, our feature-rich strategy is capable of concurrently optimizing a CMS for Quality of Context, Quality of Service, and cost efficiency, factors that dictate the usefulness of context. We demonstrate how our mechanism adapts to variations in the IoT ecosystem, context providers, and context query loads in near real-time while efficiently scaling. A large-scale autonomous vehicle management use case is tested using our prototype. Our mechanism outperforms existing benchmarks, i.e., up to 43% reduction in penalty costs and up to 99% reduction in processing and retrieval costs, while providing 27% fresher context.