Reusability of three dimensional feature based maps for automobiles

Real world roads have been mapped in 2D (google maps), but autonomous vehicles (AVs) require more information like signals or start of the tunnel. Also, the GPS signals are not present everywhere like in tunnels and their accuracy is not good enough for AVs to localize itself. For this reason, 3D ma

Project Title

Reusability of three dimensional feature based maps for automobiles

Project Area of Specialization

Computer Science

Project Summary

Real world roads have been mapped in 2D (google maps), but autonomous vehicles (AVs) require more information like signals or start of the tunnel. Also, the GPS signals are not present everywhere like in tunnels and their accuracy is not good enough for AVs to localize itself. For this reason, 3D maps are required for the vehicle to localize itself in the environment as they have very good accuracy. To complete this task, 3D maps were created by extracting the features from the sensor and then using it for localization which can be done by matching the features available in map with real time features. Our proposed model uses stereo camera as a sensor with ORB feature detection and SLAM algorithm to create a single universal 3D map of the environment that can be used for most of the environmental conditions unlike normal 3D maps that can only be used for specific environmental condition, thereby increasing the accuracy of the localization. After that, the model will be trained to do the context awareness in which we aimed to accurately identify the number of cars present before and not present now or vice versa. This whole system with certain amendments can then be used by the AV industries to help them develop their vehicles.Real world roads have been mapped in 2D (google maps), but autonomous vehicles (AVs) require more information like signals or start of the tunnel. Also, the GPS signals are not present everywhere like in tunnels and their accuracy is not good enough for AVs to localize itself. For this reason, 3D maps are required for the vehicle to localize itself in the environment as they have very good accuracy. To complete this task, 3D maps were created by extracting the features from the sensor and then using it for localization which can be done by matching the features available in map with real time features. Our proposed model uses stereo camera as a sensor with ORB feature detection and SLAM algorithm to create a single universal 3D map of the environment that can be used for most of the environmental conditions unlike normal 3D maps that can only be used for specific environmental condition, thereby increasing the accuracy of the localization. After that, the model will be trained to do the context awareness in which we aimed to accurately identify the number of cars present before and not present now or vice versa. This whole system with certain amendments can then be used by the AV industries to help them develop their vehicles.

Keywords: Autonomous vehicles, GPS, 3D maps, Localization, ORB feature detection, SLAM, Stereo camera, Context awareness

Project Objectives

• We aim to create3D maps that can be used for localization in  different seasons and time

• We aim to use these maps other than localization i.e., situational awareness e.g. to identify the number of cars present before and not present now or vice versa

Project Implementation Method

• Firstly, simulated data will be generated

• Then 3D Maps based on collected data will be made

• After checking the system if it is working on simulated data, real world data will be collected, and 3D maps will be generated for that data

• After that, all the 3D maps will be combined in such a way to have a single universal map that can be used for most of the envoironmental condtions

• Using those maps, situational awareness will be done by extracting the main car map points from the map and developing an algorithm for it

Benefits of the Project

• Accuracy of the vehicle’s position will be highly increased compared to GPS which can be used for autonomous driving

• System can also work in those areas where GPS fails to work like in tunnels

• Single map will be enough to be used in different weather conditions, without that we need specific map for specific weather condition

• System can automatically sense the environment, like how many cars are present or which traffic sign is present here and what to do when that specific sign appears on road

Technical Details of Final Deliverable

• Collection of Simulated Data using CARLA framework

• Creation of 3D Maps and using it for localization algorithm

• Collection of Real-World Data using special stereo camera and testing the localization algorithm on real world dataset

• Creation of universal 3D maps that can be used for most of the environmental conditions using ORB feature matching and SLAM algorithm

• Algorithm development for situational awareness

• Evaluation on real-world that require single board computer like jetson nano

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Education

Other Industries

IT , Telecommunication

Core Technology

Artificial Intelligence(AI)

Other Technologies

Robotics

Sustainable Development Goals

Good Health and Well-Being for People, Sustainable Cities and Communities

Required Resources

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