Self Driving Car

Project Title

Self Driving Car

Project Area of Specialization Artificial IntelligenceProject Summary

We may have heard many times in magazines and articles that there is work going on autonomous vehicles to produce cheap technology for consumers, and to provide them with vehicles that are automated and dont require any external driver assistance. Hence, we adopted this theory and tried to work on autonomous vehicle which could be cheap and fully automated, so that a normal consumer in our country could easily avail these vehicles, keeping in sight our country's meagre economic resources. So we worked on developing a prototype whose mechanism could then be put to mind to develop fully automated vehicles in future. We developed a prototype which could easily fulfill the task of lane following obstacle avoidance and traffic signal and sign detection, which can be classified as basic tasks for the initial process of development of an autonomous vehicle. In this process we took use of softwares like OpenCV and used training techniques like convolutional neural networks to gather data for our road environment and traffic signals and traffic signs. Hence doing accomplishing these tasks allowed us to go on the path of developing an autonomous vehicles. Although we fulfilled basic tasks in developing an automated vehicle but it is fully open to future work and research as future students and researches could embark on this process and could implement new features into our prototype like lane change taking u turns and equipping our automated vehicle with a larger battery and fuel reserve so that it could function for more time and cannot be confronted with the danger of lower fuel capacity.

Project Objectives

Our self drving car is setup on the basis of a few objectives, which can generally be associated with that of any initial prototype of a self driving car. Our objectives include the task of object detection through YOLO using the power of deep learning. We also covered the objective of traffic sign detection using the process of deep learning by making our own model of trained data, here we took snaps of real time traffic signs which we then trained using convilutional neural networks using the basic library of tensorflow.We then classiified the images and established a clsiifier for the data set. We then made a model of the trained images whcih served for us to enable our car to take decisions when it was confronted with a situaion of a traffic sign or any traffic signal.Furtheron, we fulfilled our objective of lane detection, which was mainly impleneted by tensorflow using the software of OpenCV. Here the main processs included to capture the video through a pi camera and then the video captured was then broken down into diferrent pixels and images, and those images were then examined, firstly the images were paseed by the process of blur then edge detection techniques were implented on the captured images, lateron, the image was divided into different pixels which enabled us to distingush our lanes from the road as the lanes were white coloured and the road was black in colour. Hence the lanes carried the highest intensity and so only the lanes with higher intensity were highlited for the vehicle hence our autonomous vehicle then followed the procedure to detect those lanes and keep its path without deviating from the road. Lastly, we had reached our objective of behaviour cloning where we gathered real time data of driving, here in our case e captured driving data of fifteen hours and hence allowed us to train our autonomous vehicle.

Project Implementation Method

This is initial Phase of GIKI toward mega projects, toward autonomus vehicles. All of the simulations of advanced algorithim that meet industry level is complete, additional to that implement on hardware using prototype of  1x1m car is also comlete. The technologies used is open source google TensorFlow library for artificial intelligence and OpenCV and languages of Python and C++. Processors used for Image processing and AI model inference are NVIDIA jetson and for controlling the motors the the microcontroller the project uses STM32 microcontroller. The camera used for video capturing is Raspberrypi V2.

Benefits of the Project

Our autonomous vehicle can greatly benefit the consumers in our country, it could be a great source of luxury as well as assistance to the general public in our country as well as the consumer class abroad. It could greatly reduce the pressure on our already meagre economic resources as we could provide cheap technology for the market in comparison to the automated vehicles in the international market. Hence it would greatly attract the consumers as the cost would have been greatly reduced other than that it would add up to the export tally of our country as we could compete in the international market with other automated companies and providing cheap automated vehicles. Further on we could add to the process of environment conservation as our automated vehicles are greatly powered by electricity and produce no harmful gases or emissions which could have adverse effects on the environment, as a result we could ease our country's pressure to develop and extract energy resources and could halt the environmental depletion  to a great extent once these automated clean energy cars are in abundant use. Moreover these automated vehicles will greatly add to the safety of our citizens as we know due to studies and results that most of the transport accidents occur due to human negligence and according to statistics about 95% of car accidents occur due to human negligence. Hence these conditions will be greatly avoided as our cars will then be fully automated and will not carry the risk of errors and mistakes. Lastly we could reduce the traffic problems in our country to a great extent as we could automated vehicles which will have no greed for space preying and overtaking as there will be no drivers involved but automated driverless vehicles which will only be instructed and will have any humanly faults and errors.

Technical Details of Final Deliverable

The project is fully functional and working on prototype of 1x1m dimensions. The car is working in real environment on local university roads and next year it will be implemented on real car. The technology of the project working is desinged for portable system that can be intergrable to any mechanical and conventional car. 

Final Deliverable of the Project HW/SW integrated systemCore Industry OthersOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Affordable and Clean Energy, Decent Work and Economic Growth, Sustainable Cities and Communities, Responsible Consumption and ProductionRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	56500
jetson nano	Equipment	1	32000	32000
STM32	Equipment	1	5000	5000
Motors	Equipment	4	1000	4000
battery	Equipment	2	1000	2000
camera	Equipment	1	6000	6000
charger	Equipment	1	2000	2000
wiring and soldering	Miscellaneous	1	500	500
track desinging	Miscellaneous	1	3000	3000
traffic and object signs	Miscellaneous	2	1000	2000