MindFeed

Finally got round to updating the live feed system with a new sexy Ajax version. No more horrid page refreshes and instead we have a smooth auto updating Google map with zoom controls and choice of hybrid, satellite, or regular map options.

Now that the mobile client is pretty much finished I decided to test it out again on a trip to London. I am glad to say it worked, though I ran out of money on my mobile phone for the live feed half way! Oops.

Some photos are below:

Today I have been tidying up code and the appearance of the mobile client to make it simple to use and provide logical feedback on what is going on - so the user knows the program hasn’t just crashed and is in fact doing something etc etc. Below are some of the mock ups of the GUI:

Today I embarked on an epic GPS recording session to really push my program to its limits by taking a train journey to London and back. Reaching a top speed of 174KMPH I can safely say that my system is robust to both fast and slow moving people!

When plotting my results using my mapping algorithm onto Google maps, I realised I needed to increase some of my upper bounds for distances between two points that were acceptable due to the fast speed of the train, but this was expected.

Accuracy was very good, keeping very well to the train track and for 2 hours of constant recording without compression the XML file size came out as 1.2Mb which when compressed can be bought down to about 100Kb.

I was also quite surprised to see GPS signals being received very nicely even inside the train away from the window which is great news too. GPS is now in a good usable state, which leaves me with the Bluetooth scanning to perfect and then I can finally develop my back end for the MindFeed website and its fancy associated services!

To ensure my recorded GPS data is as accurate as possible, I have calculated the checksums of each GPS string that is received to ensure it has arrived as expected. Any strings that fail the checksum are discarded.

Here is my client program to date.. as you can see it is progressing fairly quickly now:

Ok, so today I realised I needed to do something with regards smoothing my very noisy GPS data. To do this I combined a moving average algorithm with my own algorithm to give upper and lower bounds for tolerance levels in variation from one way point to the next.

The results are shown below:

No cleaning, original recorded data:

With my own custom algorithm:

With moving average algorithm:

Combining my algorithm with moving average algorithm: