MAE_298_Final_Project

Running environment

• gym/cartpole

Language

• Python 3.6
• Matlab for state estimation

How to start:

Install gym package for visualization

• pip install gym
• pip install filterpy

Run real time control with lqr

• python main.py -est EKF -n 500 -angle 35 -noise 1e-1 -xest 1e-1 -vest 1e-1 -thest 1e-1 -west 1e-1 --store  
  • "-est": Specify estimator, it could be EKF, UKF and KF
  • "-xest, -vest, -thest, -west": estimated process noise
  • "--store": Store data file in ./data folder, if added
  • "-n": specify running frames
  • "-noise": specify system noise (assume noises in all states are the same and constant)
  • "-angle": specify starting angles of system (other states are zeros)

Data is saved in ./data/somename.mat

• Data check in "Data check.ipynb"
• "time": time steps array
• "state_act": actual states
• "state_meas": measurement result of sensors
• "state_est": estimated state from estimator
• "inputs": inputs calculated from lqr

Plot of estimation result

• python plot_estimation.py --frames 500 --file "EKF_500.mat"
  • frames: how many frames to draw in a file
  • file: specify name of file your want to draw

System equations and LQR control

• Introduced in proposal.ipynb