Automation of Vehicle for Self Driving car

The rapidly evolving field of self-driving cars is accompanied by a journey of developments, breakthroughs, and challenges. There are profound benefits that could arise from a driverless future including safety and environmental incentives, which has led to the pursuance of this idea by many leading

Project Title

Automation of Vehicle for Self Driving car

Project Area of Specialization

Electrical/Electronic Engineering

Project Summary

The rapidly evolving field of self-driving cars is accompanied by a journey of developments, breakthroughs, and challenges. There are profound benefits that could arise from a driverless future including safety and environmental incentives, which has led to the pursuance of this idea by many leading automobile research and development groups around the world.

The concept of self-driving car is the substitution of all functions of a human driver by an intelligent control system. The primary challenge of the Final Year Project is the comprehensive vehicle modeling and analysis so that the compatibility of actuators can be ascertained. The design of autonomous control is a highly complex problem and the major areas of research include Vehicle Dynamics, Control Systems, Geometric Computer Vision and Deep learning.

The areas for active research for the Final Year Project including the objectives can be broadly classified in the following technological domains:

Perception: The linchpin for the success of the Self Driving Car will be its ability to Perceive and comprehend the environment around it, in its entirety. The requirement is fulfilled by Semantic Segmentation. To achieve semantic segmentation the most widely used AI algorithm is the Convolution Neural Network (CNN). The objective for the FYP is to collect data-sets for our road environment and perform training of neural networks to improve accuracy.

Path Planning: The path planning for Autonomous vehicles has been extensively explored and these range from the classical space configuration and curve based path planning methods, adapted for autonomous vehicles. For our project we aim to conduct the tests in a regulated environments, to fine tune our path planning algorithm, to achieve maximum accuracy and efficiency.

Control System: The controller regulates some of the states of the vehicle by sensing the current state variables (feedback) and generating suitable inputs for the plant to achieve its desired state. The objective of the project is to develop Longitudinal and Lateral control systems with disturbance regulation to achieve the desired heading and velocity.

Project Objectives

Self-Driving Cars are poised for accelerated adoption, with potential benefits such as better overall traffic management and road safety. The objective of the Final Year Project is to develop Self Driving Car in order to create a bridgehead for futuristic development of such vehicles in Pakistan, ultimately leading to self-reliance in this field. The project aims at conceptualizing autonomous driving technologies in Pakistan to bring this sector of state of art research at par with the rest of the world.

The specific scope and objectives of the Final year project are as follows:

1. Development of a prototype Self-Driving Car through conversion of an electric vehicle (done by superposing the vehicle manual control with electric actuation for throttle and steering).

2. Achieving high level of autonomy wherein the vehicle itself is capable of driving from a start point to a destination on a well-structured road with clear lane markings and minimal rush irregularities. The human driver inside the vehicle will have the supervisory control in case of emergencies.

3. Simultaneous to the vehicle development, undertake research in Artificial Intelligence and Control Algorithms to address state of the art issues pertaining to Self-Driving Cars.

4. Integrating maps, perception sensors and motion sensors for the car to be able to perceive. Reducing sensor dependencies, to camera based perception.

5. Develop a benchmark system which will facilitate subsequent research in this area for both academia and automobile industry.

 

Project Implementation Method

The implementation methodology for the project would be two dimensional. One of the aspects would be development of the Self Driving Car prototype, embedding the existing technologies and algorithms. The other one would be extension of state of the art research in this domain. Eventually both the research and development algorithms will be embedded so that a prototype of Self Driving Car with state of the art algorithms is produced as an outcome.

The development process would commence with the acquisition of an electric prototype car. The dynamic analysis of the acquired vehicle will be simulated through vehicle simulation packages like Car-sim to determine the dynamic safety limits including maximum steering angle for rollover stability and stopping distance profiles with different brake force distribution, incorporating tire models.

The selection of suitable locations for electronic drives and micro-controller will be performed. The vehicle dynamic state estimation sensors, including MEMS Inertial measurement units (IMUs) and vehicle encoders along with GPS will be installed. The IMUs will be preferably mounted at the vehicle CG for accurate state estimation. These sensors are required for determination of vehicle dynamics and response which is necessary for generation of acceleration, braking and steering inputs ensuring vehicle safety and drivability.

The actuation decisions are taken by control law which on the basis of required linear and angular positions, and velocities determines control inputs. The trajectories for the control system will be determined by the perception sensors and safety limits for the vehicle, based on traffic condition and in most of the cases, traffic regulations.

Scene recognition puts the emphasis on object detection focuses on finding object instances and their categories within a scene, typically localized using bounding-boxes performed using Computer Vision algorithms. The semantic Segmentation will be performed by the Fast SCNN algorithm, which has to be trained for our requirements, to yield efficient results within our environment.

The Path planning system determines trajectories for the low level control system. The user provides a destination location. The computer then decides how to execute the trip. The path planning task is the one to generate trajectories such that the vehicle reaches the destination ensuring safety which is incorporated through the information from maps, vehicle dynamics and the assessment of prevalent situation.

 

 

Benefits of the Project

Forecasting the way that a new technology will unfold is always a challenge. However, given the huge benefits arising from AVs, they appear all set to dominate roads in the coming decades. The developments in the mobility sector will also influence the job market, According to the surveys, more automation and control engineers and AI experts will be required in the automotive industry and at mobility service providers. The proposed project is expected to influence local job market in Pakistan which will eventually lead to increase in employment rate in the country.

Innovation

Autonomous Electric Vehicles and driving methods are highly researched and being actively developed across the globe. R&D in this area in Pakistan allows an entry and encouragement of advanced technologies in the country and therefore allows for future contribution in the development and innovation in autonomous driving technologies.

Safer Driving

According to the World Health Organization, 1.35 million people die in road crashes and another 20 – 50 million people face fatal injuries due to road traffic incidents each year. 94 percent of these fatalities are due to unsafe human driving. Human driving is prone to error and driving under conditions such as mental stress and fatigue, influence of substances and speeding make accidents more common. Autonomous driving replaces the error-bound human control of the drive with a set of trained algorithms which are less likely to make errors. Therefore, autonomous driving directly provide safety to the drivers, passengers, and pedestrians.

Traffic management

Traffic congestion and time delays are caused by inefficient human driving. These instances lead to rising levels of anxiety and stress in human beings and brings about other inconveniences. Autonomous driving techniques allow for better traffic management by reducing stop-and-go driving, understanding, and adapting to surrounding drivers. This reduces traffic congestion, saves time, and even reduces number of vehicles in use on road.

Clean and efficient driving

Autonomous driving is an efficient solution to reducing carbon footprint by the transportation industry as it reduces the amount of traffic on road and implements energy-efficient driving decisions (reduces accelerations). Moreover, as Autonomous Electric Vehicle do not use engines, carbon emissions are directly reduced. Thereby reducing pollution and not adding into the greenhouse gases emissions.

Technical Details of Final Deliverable

The subject self-driving vehicle is a one-passenger vehicle. Its power-train constitutes of a battery, power electronics, motor controller, motor, gearbox, and wheel. The power of the motor is transmitted to one of the rear tires and the rest of the tires are driven. The steering actuation architecture consists of a motor, position sensor, steering column, a rack and pinion mechanism controlling an Ackerman geometry at the front wheels.

Equipment selection for the drive train of the Autonomous Vehicle include Brush-less DC Motor controlled using an Electronic Speed Controller, and coupling of Motor with shaft through a gearbox. BLDC motors are preferred as speed control motors due to their high efficiency, silent operation, compact form, reliability, and low maintenance. The selection of Motor is finalized for a BLDC Motor, rated 1000W. The Electronic Speed controller is designed corresponding to the motor power requirements.

For Steering control, the requirements of the Stepper motor are RPM: 120 ~ 140 r/min. and Torque: 6Nm. Considering these requirements the NEMA 34 Stepper motor is selected with an encoder coupled in closed loop to ensure steps with maximum accuracy. An encoder is required to keep track of manual steering operation to allow manual control from Driver.

The vehicle will be powered by rechargeable batteries. Lithium-ion batteries will be used as they have a high power-to-weight ratio, high energy efficiency and performance. The requirements of the battery are 48V and 30Ah, which will be achieved through 13s 10p cell configuration.

For perception, cameras will be used. The Stereo Cameras provides accurate depth sensing with a flexible range between 0.5 to 18 meters. The Fish-Eye lens camera covers the entire side view and also supports the other cameras, leaving very little blind spot. The sensor requirements include VL53L0X Time-of-Flight Distance Sensor, and u-blox NEO-M8N GPS sensor.

The vehicle constitutes of multiple subsystems, controllers, and different interfaces eg. Sensors and actuators. So in order to efficiently manage them these subsystems are connected to our main ECU through CAN-BUS topology. Every subsystem consists of ATMEGA328p micro-controller, along with relevant sensors throughout the vehicle. For CAN BUS we are using MCP2515 CAN controller, connected to our micro-controller through SPI interface. The Electric and Autonomous components and algorithms constitute our final deliverables.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Transportation

Other Industries

Education , IT , Energy

Core Technology

Robotics

Other Technologies

Artificial Intelligence(AI)

Sustainable Development Goals

Industry, Innovation and Infrastructure, Sustainable Cities and Communities

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com