All the four of us as classmates pursuing a masters degree in computer science worked in variety of domains but all of us were always excited about working for energy consumption and sustainable development, some way or the other.
What it does
Our proposed work attempts to give solutions to two major purposes. One of them was : Given the street light setting employed by ESSL, our proposed work actually finds out the exact street light that has gone at fault with just the already existing monitoring system.
Secondly : We have a complaint management problem. We go on to forecast the number of complaints that would occur on a given time in a given location.
How we built it
First Challenge : In Challenge 1 we added high and different resistance to every lamp in parallel and calculated power. We also take the fact that maximum possible fault configuration can be 2^n, where n is no. of street light. now there are two approach to solve the problem. one is by 'additive factoring' another by deep learning model. Computation time required for deep learning will less as compared to additive method. Once model is trained we need not to train it again. by the virtue of the design the proposed work is user centric because when any lamp faults the power usage of that particular faulty lamp becomes higher which give incentives to the lamp provider to quickly fix the faulty lamp.
Second Challenge : We go on to treat it as time series prediction problem. From the given data, that had information about the complaints lodged in last six months. We build a deep neural network using LSTM units, that will forecast the number of complaints that will occur at a given time, given the past patterns. Earlier, we decided to build separate models for separate states such that more and better quality of information could be captured, the whole working is smooth and viable. And with these, we can put the resources that are basically the response teams against the complaints, put to use in an optimal manner.
Challenges we ran into
There were multiple challenges and road block that we hit during our last 26+ hours of journey. The problems were just to find an optimal way of doing it and not what to do. In other words, we were well informed about the technology that we had to put into use to come up with this solution we just had to create the road map and that happened to be the most challenging part. For another thing, we all are technical people, coming from computer science background and hence, we still are facing difficulty in generating a revenue model out of our work.
Accomplishments that we're proud of
Accomplishments that we are the most proud of must we the way we worked as a unit, coming from almost very different interests, we talked to each other, we went to the mentors and asked the right questions. Also, time management. And most importantly the fact that we could sit on a round table, brain storm and move towards achieving the goal.
What we learned
We did not learn anything new in technology bit, i.e. the programming language, the libraries or the API's. We all were well familiar with all those we just learnt non-objective things like, team work, time management, asking the right questions at the right time to right people and collaborate our ideas.
What's next for AARCH64
AARCH64 has a vision to soon turn into an independent firm, establish a research lab and develop tangible IPs. Covering right from the IoT domain to low level architectures. From the simple analytics to products and solutions with advanced Artificial Intelligence.