Home > discussions, Ideas, projects, Supervisor Meetings > PhD Direction Meeting 13 Nov 2008

PhD Direction Meeting 13 Nov 2008

November 13th, 2008 Leave a comment Go to comments

Met with Simon and Aaron this morning to talk about the direction of my PhD 🙂

I introduced the city sensing idea that Paddy and I had talked about – and that I had previously sent to Simon to comment on.

I also started by saying that I am interested in the problem of Delay Tolerant Networks.

Firstly, we tried to sketch out the problem to be solved – I could not really describe it, so Aaron suggested that one of the problems is that of infrastructure – i.e. it is very expensive to install and use. For example, is every phone in Dublin was reporting sensor data, it would firstly swamp the network, and secondly be prohibitively expensive.

Simon also suggested that it was a problem of coverage – in areas where there is no infrastructure, how to do you report sensor data?

Aaron and Simon stressed that it was important to relate this problem to the real world by giving real examples – such as figures for infrastructure cost – data usage and network utilisation. But if we cannot get figures for this, we can make estimates based on what we do know.

Aaron mentioned an experiment he did when in Australia with ?(Dan)? which tested bluetooth hand-off times – and found that the handshake time was very long – and too long for passing nodes to exchange data. Simon and Aaron therefore suggested that a low level protocol could be designed, that if we assume it existed – could form the basis from which my contribution works.

This would mean that any system based on this, would have to be simulated, as the protocol itself would not be implemented for real (not by me anyway). Ideally however, I would have a system running on every phone in Dublin from now until when I analyse the data, however, this is not going to happen. But having it run on 10 phones would be nice, on 100 would be difficult but better, on 1000 phones, very very hard but very nice.

However, at this point, my contribution is unclear – so we discussed some ideas:

  • Micro messaging of data, small packets sent until out of range, making it as efficient as possible, some data wont get sent, but some will.
  • Reporting that things are normal, but not sending the data, as it can be derived that if there is nothing abnormal, then we wont send data we just ACK that everything is ok. – I suggested that this may be similar to Erasure coding which xors messages against previous known messages to extract the new message.
  • Perhaps we could utilise a mechanism for piggy-backing data over existing protocols – such as Bluetooth Friendly names, or spoofed MAC addresses, or on ARP requests.

We discussed simulations, and Aaron noted that Brendan has cluster access for processing large amounts of data – and also mentioned OmNet for running simulations.

Aaron mentioned that when they wrote their own simulation, it was deemed a bad idea, because there is no way to evaluate it, so suggested that we should not write out own simulator for anything.

I mentioned that a paper I had read had used information from wireless access points use (CRAWDAD) to simulate a network of people, and mentioned that I had contacted IT services at UCD to see what data they keep about access point use, and whether we can use it. Aaron suggested that I email him and Simon with my request and they will see if they get access to the data.

Simon mentioned Bill Hilliers book – space is the machine – which is worth dipping into when considering how to simulate city environments.

Aaron said that his student ?Dan? had taken a real un-related dataset, and synthesised a relevant data set from it (iSpy search dataset).

Another dataset that is apparently available is live data from traffic lights in Dublin, which shows the position of the lights, the number of cars queueing and is available live onlin.  – I wil email Kevin McCarthy who may know who has information about this.

I suggested that we could get traces of data from second life, which may give a good dataset for simulating the movement of people around a geographic area. This has inherent problems, but might be interesting.

I Suggested that there was (based on a survey paper by Zhensheng Zhang 2006 [1]) a research area for profiling nodes in a network to predict their future movements, which prompted:

Simon gave an example of a system used at Reading university where they tracked people in a building using RFID tags, and found that people generally had very predictable routines.

Simon also mentioned a paper by Sandy Pentland about familiar strangers – the people you come into contact with regularly, that you don’t necessarily know – these people could be used (discretely) to send messages to people they may come into contact with.

Aaron introduced a Clarity Project called Sense in the City, or more specifically GOYA which he thinks I may be able to become involved with, and he suggested that I email Paddy and Barry to say that I would like the slides from the CLARITY away day, and that I would like to find out more information about the project it becomes a reality – but that I shouldn’t spend too much time on finding out about it at the moment.

The project envisages using a network of bicycles which are installed with sensors, that people use to get around the city. He suggested that it is a good idea to have a case study – such as this – which tests my system in the real world with interesting results.

He also mentioned the idea of Healthy Spaces – urban parks where people walk around – may not be obvious on the map – but in return for generating interesting sensor data, we can map out where these places are – and perhaps contribute to the health credits idea (discount on Health insurance).

We identified some next actions:

  • Matt: Document: define the problem, define the motivation
  • Matt: Email paddy + Barry for slides
  • Matt: Email Simon + Aaron about wireless access points log access
  • Matt: Email Kevin about traffic information
  1. No comments yet.
  1. No trackbacks yet.