Nitin J Sanket
Path Planning and control of a quadrotor to follow a specified trajectory
Aim:
Implement the trajectory generation and control strategy on a quadrotor to follow a specified trajectory.
Description of the Approach:
Step 1: Path planning using A*
In this step, a simple A* based serach strategy was used to a plan a path from the start position to the goal position. The assumption here is that we had a complete map of the environment with the obstacles.
Step 2: Trajectory generation
Quintic spline based trajectory was generated using the output of A* planned path.
Step 3: Linear Controller for the quadrotor
In this step, a linear controller based on hover state was desined using PID. The current quadrotor pose was obtained from the Vicon motion capture system.
Step 4: The whole system was tested in simulation and on a real quadrotor.
Vision Based Quadrotor Control
Aim:
Control a quadrotor using pose estimation from a camera on board.
Description of the Approach:
Step 1: The quadrotor pose is calculated using April Tags on the ground whose position in the world frame is known. The projection equation and PnP is used to estimate the pose.
Step 2: The linear and angular velocities are computed using optical flow and 3-point RANSAC is used to reject outliers.
Step 3: Both the position and velocity estimates are fed into an extended kalman filter to achieve the required control policy.
Multirobot coordination
Aim:
Assume we have a set of N robots at given start locations that need to complete (navigate to) M tasks
at given goal locations. Further, assume that the robots are interchangeable, that is, it doesn't matter which
robot visits which goal location, as long as all goal locations are visited. The problem is to assign robots to
goals and then generate trajectories that will navigate robots to their assigned goals in an optimal and safe
manner.
Description of the Approach:
This project used C-CAPT and D-CAPT to solve this problem.