Metaswitch Networks
Contact: Beth Gorman <Beth.Gorman@metaswitch.com>
Potential client: John Palombo <John.Palombo@metaswitch.com>
1. Beer Goggles
Augmented reality project for Google Glass that displays the distance to the nearest pub and shows the direction to go to get there. Obviously, this could be made more general to provide directions to other types of destination, either at a street level, or for example giving directions to a user in an unfamiliar building, but the name wouldn’t work then!
We’d provide one pair of Google Glass glasses (to be returned!).
Feedback: We'd like to include some Google Glass apps this year, and I'm sure students would be interested. However, we have done quite a few navigation apps in the past, so I'd like to extend this one (also slightly concerned that we had to cancel a project called "wine goggles" last year after lack of interest from students, who perhaps are not so interested in alcohol as we might assume).
2. VoIP Network Quality Tester
VoIP and Video calls are very sensitive to latency, jitter and packet loss in IP networks, but there aren’t many good end-user tools available to determine how good a network is. This project creates a probe that users can run on a standard end user device (e.g. desktop, tablet or smart phone) that communicates with a remote server and provides an indication of how good the network is and predicts the quality of calls through it. Ease of use is essential – both to install/run and presenting the results in a user friendly manner. Idea could be extended for example to crowd-sourcing across many users in order to build a much wider picture of the network, or to predict quality of a gaming connection.
Feedback: Very timely. However, students may not have suitable signal processing skills. Would this use separate audio hardware to inject and capture controlled signals, or rely on intercept of system audio? Would capture and analysis of video quality be an alternative?
We are increasingly deploying our telephony systems in cloud environments as that allows us to increase and decrease capacity elastically in a way that has never before been possible. In the past, the only way to cope with peaks in demand (such as during X Factor voting) was by engineering in lots of capacity which of course adds expense. One of the challenges of an elastic system, though, is that it can take of the order of ten minutes to bring a new resource on-line, so judging when to increase capacity is not straightforward. If you bring the resource on-line too early the carrier faces increased costs, but if you bring the resource on-line too late calls will fail and the carrier faces lost call revenue and upset customers. To make matters more interesting, resource cost may not be fixed, but may have steps or vary by demand and/or time of day.
This project would define an algorithm (or algorithms) to decide when to add and remove (‘orchestrate’) resources in the cloud to balance the competing pressures of cost and customer service.
We will provide datasets that can be used to train the system and to test how it behaves with new data.
Feedback: An interesting challenge, but I don't think it would be easy to divide this among a team of six. Is there some other way that a team might create a demonstrator based on the data sets that you have?
4. Raspberry Pi 3D Scanner or Lidar
[Not sure if this one is possible]. Develop algorithms that run on the GPU processor on the Raspberry Pi that can be used to convert it into a low cost 3D Scanner or Lidar.
Feedback: We have done 3D scanners in the past, and have also done projects based on Raspberry Pi GPU. From those experiences, the results have been a little disappointing - the GPU is quite closed, and students don't have the hardware skills to create an accurate scan.
I did have a related idea that is feasible, that we don't have a client for at present:
Use a 2D scanner to capture the section of an unknown section of PVC or aluminium extrusion. Match automatically against online catalogues to order. If not found, 3D print to required length.
