Source author record

Philip Valencia

Philip Valencia appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

cs.CY Networking and Internet Architecture Neural and Evolutionary Computing

Catalog footprint

What is connected

2works

3topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2015arXiv

Delay-Tolerant Networking for Long-Term Animal Tracking

Enabling Internet connectivity for mobile objects that do not have a permanent home or regular movements is a challenge due to their varying energy budget, intermittent wireless connectivity, and inaccessibility. We present a hardware and software framework that offers robust data collection, adaptive execution of sensing tasks, and flexible remote reconfiguration of devices deployed on nomadic mobile objects such as animals. The framework addresses the overall complexity through a multi-tier architecture with low tier devices operating on a tight energy harvesting budget and high tier cloud services offering seamless delay-tolerant presentation of data to end users. Based on our multi-year experience of applying this framework to animal tracking and monitoring applications, we present the main challenges that we have encountered, the design of software building blocks that address these challenges, and examples of the data we collected on flying foxes.

preprint2014arXiv

Genetic Programming for Smart Phone Personalisation

Personalisation in smart phones requires adaptability to dynamic context based on user mobility, application usage and sensor inputs. Current personalisation approaches, which rely on static logic that is developed a priori, do not provide sufficient adaptability to dynamic and unexpected context. This paper proposes genetic programming (GP), which can evolve program logic in realtime, as an online learning method to deal with the highly dynamic context in smart phone personalisation. We introduce the concept of collaborative smart phone personalisation through the GP Island Model, in order to exploit shared context among co-located phone users and reduce convergence time. We implement these concepts on real smartphones to demonstrate the capability of personalisation through GP and to explore the benefits of the Island Model. Our empirical evaluations on two example applications confirm that the Island Model can reduce convergence time by up to two-thirds over standalone GP personalisation.