An implementation of iLQR for trajectory synthesis and control. Use finite difference to approximate gradients and hessians if they are not provided. Also support automatic differentiation with numpy from jax. Include an inverted pendulum example as the test case.

Dependencies:

Numpy

Matplotlib (Only for the test)

jax (Only for automatic differentiation)

pytorch (Only for learning-based MPC test)