Computer Laboratory > Research > Dr Salman Taherian

The following projects (for 2011/2012) are centered on real world applications with some elements of analytics and software implementation.

---

Project 1: Optimal use of rechargeable battery units for serving a building’s energy (electricity) consumption

The emerging smart grid and smart meter devices measure aggregate electricity consumption at frequent intervals (half-hourly measurements). Fluctuations in the data reflect the increased or decreased usage of services – some services (e.g. boiling a kettle) are short lived while others (e.g. electric air con) are long-lived and are more noticeable in the aggregate data. Green energy sources (e.g. solar panels) can charge on-site batteries and this limited amount of energy can be used to serve short-lived services; thereby creating a lesser dynamic energy consumption profile for demand prediction and optimal power-level allocation.

This project looks at real energy data from a few governmental buildings and aims to decompose the trend into high frequency and low frequency components for association with the mentioned short-lived and long-lived services. An evaluation framework must be defined to judge how well this decomposition is performed; metrics of interest may be a) the level of exploitation from a predefined rechargeable battery reserve and b) the level of reduction in the peak demand.

Project 2: Automated on-my-way android application

When we’re late for a meeting, we call/msg the other party with some estimated time of arrival (e.g. I’ll be there in 10 minutes!). Smart phones now have accelerometers that can detect change in movement, and GPS sensors that can accurately position a person in an area; along with many other useful information.

This project will be looking to automate the above process in two ways a) estimating a departure time from assessing a series of sensory information and b) estimating an arrival time at a given destination based on speed of movement/progress – you could even tap into Google services if you wish to enhance these estimations. The end goal is to produce an Android application which when given a target meeting point (location), time, and contact detail (for the other party), it can monitor your current position and behavior and detect when you’re likely to arrive at the destination (early or late) and send this info in the form of a txt msg to the other party.

Project 3: Personalized resource allocation for Windows processes

At present, PCs (in the very traditional sense of Personal Computers) are very much under-utilized. This has motivated concepts of cloud computing, thin-clients, netbooks, etc where applications are externally hosted and resources are dynamically allocated. On-demand resource allocation, however, is costly. An alternative approach is to estimate/predict the level of resources that a particular user or application may need and allocate that ahead of time; the aim is to reduce the user’s under or over-utilization of his/her allocated resources. Past resource usage information is our most valuable asset in performing this prediction.

This project looks to implement a Windows Process monitoring application that would extract resource usage information (relating to each user application) at frequent intervals from the operating system. This information will be analyzed and key input signals will be identified that can explain the observed resource usage level. These input signals will then be used as part of a prediction model to estimate resource usage levels for every user and application based on his/her interactions with the application.

---

Drop me (st344 at cl.cam.ac.uk) a line if you're interested in any of these areas or the enlisted projects.