This project implements a comprehensive system for quadrotor autonomy, including state estimation, path planning, trajectory optimization, and control, all developed from the ground up.
-
Utilized A* algorithm with engineering optimizations:
- Diagonal heuristic
- Cross tie breaker
-
Achieved 20x planning speed improvement in 3D grid maps
-
Future improvements:
- Integration of dynamic model with state-space planning (e.g., State Lattice Search, Kinodynamic RRT*, Hybrid A*)
-
Implemented minimum snap trajectory optimization
-
Based on paths generated by A* algorithm
-
Solves kinodynamic constraints with boundary conditions in Cartesian space
- Implemented quaternion-based Unscented Kalman Filter (UKF) and complementary filter
- Achieved 20% improved efficiency compared to rotation matrix implementations
- Deployed complementary filter on onboard IMU due to computational resource constraints
Esitmation results of the quadrotor:
Gound truth:
- Developed custom control algorithms for quadrotor stabilization and trajectory following
- Utilized PID controllers for attitude and position control.
Orientation:
Position:
Commands:
