Autonomous VIO-based Quadrotor

This project is the simulation of Stereo Visual Inertial Odometry (VIO) on a quadrotor modeled from CrazyFlie 2.0. In the simulation, the quadrotor is set up with an on-board IMU and a stereo camera rig.

Pipeline

• A* algorithm for shortest path searching and minimum snap for trajectory smoothing.
• Geometric Nonlinear Controller for trajectory tracking.
• RANSAC for robust feature matching given stereo correspondences to recover camera pose.
• Error State Kalman Filter to acquire accurate quadrotor pose estimation.

Run

• Setup the environment:

  pip install .

• Run VIO simulation:

  python sandbox.py

• New maps can be customized through JSON file and saved under proj/util/.

Examples

• Shortest path generated by A* algorithm, pruned waypoints and smoothed trajectory:

• Trajectoy tracking: