sushain.com

Everyday, we travel from our homes to our workplaces, jostling and cringing about the congested roads and unorganized traffic, coupled with the unpredictable weather and several other factors that make it very hard to commute. Everyone shares such emotions, whether using a self-owned vehicle or a public transport.

Inspired, rather provoked, by this problem, we came up with an idea to build a solution, which would help solve this problem to some extent. The Solution helps commuters identify the best path on the fly as they’re traveling, given a source and destination. The system calculates the resulting path based on live traffic and weather-related feed data from an external source and display it on a geographical map in real-time. Best path, in this case, is calculated holistically; mainly by analyzing a specific live, textual RSS traffic-feed, provided by Yahoo’s Traffic API and then coupling the result with a live RSS weather or current events’ feed. Finally, these results are displayed on any freely available Map Service, like the one provided by Google, so that the user can easily view the path that he/she needs to follow.

To start with, the solution needs to be restricted to those areas, which are covered by Yahoo’s Traffic API. By doing some fundamental feasibility analysis, it was found to be practically available for the southern regions of New York and New Jersey. The assumption was that once this proof of concept is build, the dependence on the Yahoo service can be eliminated by putting a live video-feed analyzer component that doesn’t depend on the text-based feeds, leading to a generically applicable solution.

The major component of the system, the Feed Analyzer, is basically an intelligent component that acts as an Expert System. It parses the incoming XML feed and assigns severity based on the presence of some keywords, simultaneously learning and training itself to classify the future feeds in specific severity classes. This results in an analysis, which is monotonically increasing in the nature of precision. Finally, the path with least overall severity is characterized as the best path.

The other key component is a Pathfinding algorithm, which works on a connected graph, consisting of locations and roads (as nodes and points), respectively and finds the set of all possible paths between a source and a destination. After doing this, the output from the feed-analyzer is utilized to assign severities to each of these paths. Again, this algorithmic ‘manual’ component can be made redundant if there is an externally available, global and generic service or an extension to an already available service API, like Google Maps, that helps in finding all possible genuine and physical routes given two points, located by their longitude and latitudes.

Finally, as the feed analyzer is an independent component that takes XMLized feeds as input, there can be multiple and different inputs to it, or it can utilize an output from Yahoo pipes that is produced after merging many different and relevant feeds, thus increasing the possible relevance and practicality of the overall solution.

Here is a demo of the resulting proof of concept: -

Artificial Intelligence, Demonstration Artificial Intelligence, Javascript, Mashup, neural networks, ruby, traffic, vipi