title-abstract.txt

Title: 
Autonomous Mapping, Localization and Navigation using Computer Vision as well as Tuning of Camera

Abstract:
One of the main tools used in autonomous mapping and navigation is a 3D Lidar. A 3D Lidar provides various advantages. It is not sensitive to light conditions, 
it can detect color through reflective channels, it has a complete 360 degree view of the environment and does not require any ”learning” to detect obstacles. 
One can use the reflective channel to detect the color of lanes as well as avoid obstacles. The pointcloud information from the Lidar can also easily enable 
mapping and localization as the vehicle will know where it is at all points. It is easy to see why so many large scale autonomous vehicle units invest in 
expensive and bulky Lidars. However, this is not accessible to all due to it’s price. A camera (even depth) is much more affordable. However it comes with 
it’s own slew of disadvantages. It can see color but programming for the color is hard due to varying light conditions. Unless you use multiple cameras you 
often can’t see all around you. These factors together are a hindrance to autonomous navigation. We thus aim to demonstrate 3 goals:

1) Mapping and Localization with a single camera and other sensory information using RTABMAP SLAM algorithm
2) Obstacle avoidance and lane following with a single camera using Facebook AI Detectron2 Deep Learning and ROS
3) Tuning of the camera to be less sensitive to varying light conditions using ROS rqt_reconfigure