Manipal’s First Self Driving Car!
On 24th March 2014, Mahindra’s “Spark the Rise” challenged the brightest minds of India to build a driverless car for Indian roads. The members of Project Manas crossed innumerable hurdles and challenges to make it to the top 13 among 259 participating teams. The purpose of this event is to revolutionize driving on Indian roads, thereby transforming lives with the promise of positive changes in many forms: from decreased road mortality rates and driver fatigue, to increased environmental friendliness that makes transportation more inclusive for people of all abilities. Kumar Ranjan and Roshan Prakash, the two geniuses who took up this challenge competently, brought immense pride to our college. MTTN had the privilege to meet the people who were responsible for taking the initiative, knowing how it all began, and making it what it is today. Here’s what they had to say.
MTTN: Whatever little we have read about this, Google has a self-driven car, Tesla has an electric car, so what is different about your project? Are you aiming for a combination of both or are you implementing your own ideas?
PM: Mahindra will be sending us a Reva e2o car. After some days we will be getting the e2o car from Mahindra itself. For now, we are working on electric cars. So you can say it is a combination of both.
MTTN: What made you come up with this project?
PM: It was an event by Mahindra Rise- it is a one million dollar challenge. Suddenly, this challenge comes up and our team leader Kumar Ranjan wishes to take part in this project. That is how it all begun. Since then we have been working on the technical and management part and yes we have come a long way.
MTTN: How many rounds have you completed?
PM: We have completed the first round- and it had three parts. The first part involved presenting a general idea of what we were planning to implement, basically “what is a driverless car?”, and how we propose to build it. We had to submit an approach note to Mahindra describing what technology we have. The next stage, which we have already begun, is the prototyping phase.
MTTN: We saw the conception of Wall-E on your website. So since then, have there been any changes in Wall-E?
PM: Yeah we have added innumerable new features. Initially it was not capable of certain features, but a lot of development has taken place in the software part hence. The main development is related to stereo and LIDAR. It is now capable of creating the 3D picture (3D point cloud of the picture). Also, based on LIDAR if a person comes in front of this bot suddenly, it stops. Hence it uses these odometric techniques to localize itself in the environment. We also have a stereo camera to identify different obstacles. Besides, there is a road detection function too-it helps detect roads, signs and so on.
There are a few videos on YouTube that we have submitted to Mahindra. You can see the videos like the ones on pedestrian detection, phase detection and so on. We wanted to implement it on this system so that it can be more robust and can use its own intelligence to make its own decisions. For example, if a pedestrian or a car or any dynamic object comes in front of it, it stops.
MTTN: Suppose there are two cars, both automated. Do they actually interact with each other?
PM: We are working on that. Mahindra gave us deliverables. The deliverables contain certain challenges. One of the challenges involve solving the issue of communication between cars if there are more than one. This is known as M to M- Machine to Machine interaction. They are also applying a similar concept to track the vehicles. As far as implementation is concerned, we have already implemented a J2EE server.
MTTN: So what particular aspect hinders the implementation of this? Is there any particular aspect that hinders the implementation of this vehicle?
PM: As with any project, the hindrances are many. One by one we are tackling them and we have devised our own matrix for this. One of them was a fluctuation in readings- say stereo cams. If the bot is moving at a relatively high speed and there is no shock absorbing mechanism, the camera features vibrating images. We are yet to find a way to stabilize this.
MTTN: Does the camera detect a relative motion?
PM: Yes, it is called the Global Motion Detection. It was a problem and as of now we have not found a solution to completely solve it. For the time-being, our algorithms are working without this feature but it is necessary. Moreover as in any sign detection procedure, the bot should detect and then stop. We are trying to bring about this sort of parallel computing. It is something we haven’t implemented yet, so we need to tackle that too.
MTTN: Are these cars going to be remotely controlled by servers?
PM: Servers are not for controlling the cars. Servers are only for communication. It gets information from the car and transmits big data. All the data from different cars are uploaded on the server and the server has to analyze all the data and provide some feedback to the cars. Like if it encounters some traffic -huge amount of traffic in Bangalore- then the central server will manage all this data. For example, Swarm Intelligence. They will figure which route to opt for to reach the destination quickly, as well as manage excess traffic. It is not that you can control the server from the car, you can just provide information so that you can do something which is very optimal.
MTTN: What if someone were to hack into the server? What kind of security is there to inhibit hacking into the server?
PM: J2EE- Java Enterprise has its own security features. Java is well known for the security it provides. Apart from that, secure protocols are used. And since security is a big concern, we ourselves don’t know what bugs we might find. Besides the economy is a big issue. We don’t want to implement security features from scratch. We go for third parties which can provide us security soft wares.
MTTN: You are aiming to create the first automated car. So how are you planning to tailor to fit the needs of an average Indian? Because it will of course be expensive.
PM: By commercial, we mean government sectors- such as buses and ambulances. The car will definitely be costly. It is yet to be tested too.
The challenge revolves around making a driverless car for Indian conditions. Google has already made a car, and many others are coming up, hence the challenge lies in making cars for Indian conditions. We need to take into account economy, and terrible road conditions. One of the steps that we have taken is to not to go for proprietary soft wares. Our entire system is based on an open source. There are no licensing issues, no patents, and no royalty costs in that respect. And as far as production costs are considered, it is not under our control. Yet, we will try to get the best sensors and utilize it in the cheapest way possible.
MTTN: Can you tell us more about the involvement of Firexit software in your project?
PM: Actually when the project started, there was no AI subsystem in our project- it was lacking the Artificial Intelligence subsystem. So Roshan and I worked on the automation. I called Shubham- asked him if he was interested in this project. He was already involved with Firexit software and he said Firexit would like to contribute- so it is a kind of collaboration. I was a researcher there for Computer Vision, and Augmented Reality so Vatsal wished to collaborate too as it was a new startup. The core AI routines are contributed by Firexit. But with new teams the algorithms will belong to Project Manas itself. However, we owe a lot to Firexit to have given us the head start.
MTTN: What quality do you look for in the new recruits?
PM: We want them to know the basic things. In first year they don’t know much about their subjects. They need to know about Arduino, relays, simple sensors. If they are into AI they should know about AI and the basics of AI. They should have good knowledge of coding. A basic knowledge of C, C++ would not suffice, they need to delve deeper into their programming skills.
MTTN: Of course the knowledge we gain in first year is definitely not enough to work for The Project Manas team. So what extra knowledge is required and from where do we get this extra knowledge?
PM: Some students start working hard from their first year itself. Sometimes having worked on certain projects using Arduino, having participated in Tech Tatva, indeed helps in gaining the required knowledge. We also motivate them to watch lectures and try to solve problems of courses in Coursera.
MTTN: What is the major source of funding?
PM: The college has funded us, and we are also looking for sponsorships. The director was interested in funding more towards our project, and Mr Ranjan Pai also wished to invest and give more space to this project. Dr Ashalata maam is the mentor for our team.
MTTN: Any last message you would like to give our readers?
PM: From what we have seen there is a shortage of girl coders in our branch. It would be better to have more girl coders who are profound in their knowledge of coding and have interest in our subsystems. So we would like to send a message to the girls to take interest in coding and venture forth to join our team.
Founder- Kumar Ranjan
Co-Founder and Team Leader- Roshan Prakash
AI Subsystem Head- Shubham Verma
Technical members (AI) – Kaushik Nath
Automation and Sensing – Akash Tyagi
Mechanical- Rohit Singh
Management- Mansi Goyal
To know more about the project Manas team click here
To know more about the event by Mahindra rise click here
As said to: Ananya Panda and Prattusha Mukhopadhyay.
Photography by: Samyukta Pingaali.