Trajectory-Optimization

This repo contains trajectory-optimisation on some basic systems (ex: pendulum,cartpole ,quadrotors and manipulators). Will be implementing algorithms like finite horizon LQR, iLQR , AL-ilQR,Box DDP and DIRCOL.

Installation details:

git clone https://github.com/yaswanth1701/Trajectory-Optimization.git
cd Trajectory-Optimization
git submodule update --init --recursive

• Make sure to run gym-pybullet-drones inside the conda environment (for details click here).

  Demo:

  video (credits: ppap36)

Algorithms :

• [X] Infinite horizon LQR:

with external disturbance(0-50N) -

• [X] Iterative LQR :

• [X] Iterative LQR (with line search) :

• [X] Iterative LQR (with finite-horizon LQR for trajectory tracking) :

  

• [x] Infinite LQR (with quaternion as state):

• [X] MPC using DIRCOL :

for code refer here.

Current environments:

• [x] Cartpole (gym)
• [x] Pendulum (gym)
• [x] Quadrotor (PyBullet)
• [x] Turtlebot3 (gazebo)
• [ ] 5 link Biped (matplotlib)

  References:

• Notes
• Advanced robotics by Pieter Abbeel (iLQR)
• Optimal control (CMU AL-iLQR)
• Underactuated robotics (For non linear dynamics)
• Control-Limited Differential Dynamic Programming (Box DDP)
• really good explanation for DDP