Click Here for Abhyast Phase VI Problem Statement
Phase VI has been started with the problem statement of implementing a robotic system that would autonomously map an unknown environment localizing the suspicious objects and providing this information to a ground station where a manual controller will wirelessly direct a ground vehicle to pick that object(s) and take it to a safe zone, thus avoiding the loss of human lives. The localization and mapping of the area would be done through an autonomous swarm of aerial vehicles. The ground station will utilize this data to identify the suspicious object(s) and plan an optimal path over which the controller will maneuver the ground robot using the live feed obtained from the cameras onboard, to acheive the required task. The phase started successfully and we are making progress on the task.
Point Cloud Generation: Point clouds has been successfully been generated using Laser Scan data from laser scanners and transformation data from IMU sensors.
Visual Odometry: Using a stereo setup of two logitech c310 webcam , we have been able to extract the motion of camera in the form of transformation matrices using opencv and nister-5-point library. we will use this information in building 3d map of the environment and doing Visual SLAM
Optic Flow: Using imu data from ARDUCOPTER and video feed from logitech webcam c310 , we have been able to extract the velocity(linear and angular) of camera looking at a planar surface !! we are working on extending this concept to a general structure removing the restriction of planar scene.
Actuator Control: Codes for actuator control has been successfully tested and would be optimised for stable and autonomous flight.
3D reconstruction using Monocular Camera: Point Clouds were successfully generated and 3D reconstruction was implemented using moving Monocular Camera but the results were very slow and ineffiecient.