acrobot_model_free_DDP.py

'''
copyright @ Karthikeya S Parunandi - karthikeyasharma91@gmail.com
Model free DDP method with a 6-link acrobot experiment in MuJoCo simulator.

Date: July 6, 2019
'''
#!/usr/bin/env python

import numpy as np
from model_free_DDP import DDP
import time
from mujoco_py import load_model_from_path, MjSim, MjViewer
from ltv_sys_id import ltv_sys_id_class
from params.acrobot_params import *
import os


class model_free_acrobot_DDP(DDP, ltv_sys_id_class):
	
	def __init__(self, initial_state, final_state, MODEL_XML, alpha, horizon, state_dimemsion, control_dimension, Q, Q_final, R):
		
		'''
			Declare the matrices associated with the cost function
		'''
		self.Q = Q
		self.Q_final = Q_final
		self.R = R

		DDP.__init__(self, MODEL_XML, state_dimemsion, control_dimension, alpha, horizon, initial_state, final_state)
		ltv_sys_id_class.__init__(self, MODEL_XML, state_dimemsion, control_dimension, ctrl_state_freq_ratio, n_samples=feedback_samples_no)


	def state_output(self, state):
		'''
			Given a state in terms of Mujoco's MjSimState format, extract the state as a numpy array, after some preprocessing
		'''
		#print(self.angle_normalize(state.qpos),"\n")
		return np.concatenate([self.angle_normalize(state.qpos), state.qvel]).reshape(-1, 1)

	def angle_normalize(self, x):
		'''
		Function to normalize the pendulum's angle from [0, Inf] to [-np.pi, np.pi]
		'''
		return -((-x+np.pi) % (2*np.pi)) + np.pi

	def cost(self, x, u):
		'''
			Incremental cost in terms of state and controls
		'''
		return (((x - self.X_g).T @ self.Q) @ (x - self.X_g)) + (((u.T) @ self.R) @ u)
	
	def cost_final(self, x):
		'''
			Cost in terms of state at the terminal time-step
		'''
		return (((x - self.X_g).T @ self.Q_final) @ (x - self.X_g)) 

	def initialize_traj(self, path=None):

		'''
		Initial guess for the nominal trajectory by default is produced by zero controls
		'''
		if path is None:
			
			for t in range(0, self.N):
				self.U_p[t] = np.random.normal(0, nominal_init_stddev, (self.n_u, 1))	#np.random.normal(0, 0.01, self.n_u).reshape(self.n_u,1)#DM(array[t, 4:6])
			
			np.copyto(self.U_p_temp, self.U_p)
			
			self.forward_pass_sim()
			
			np.copyto(self.X_p_temp, self.X_p)

		else:

			array = np.loadtxt('../rrt_path.txt')
			### INCOMPLETE


if __name__=="__main__":

	# Path of the model file
	path_to_model_free_DDP = "/home/karthikeya/Documents/research/model_free_DDP"
	MODEL_XML = path_to_model_free_DDP + "/models/acrobot.xml"
	path_to_exp = path_to_model_free_DDP + "/experiments/acrobot/exp_10"

	path_to_file = path_to_exp + "/acrobot_policy.txt"
	training_cost_data_file = path_to_exp + "/training_cost_data.txt"
	path_to_data = path_to_exp + "/acrobot_D2C_DDP_data.txt"

	
	# Declare other parameters associated with the problem statement
	
	alpha = .3
	n_iterations = 90


	# Declare the initial state and the final state in the problem
	initial_state = np.array([[0], [0], [0], [0]])
	final_state = np.array([[np.pi], [0], [0], [0]])

	# Initiate the above class that contains objects specific to this problem
	acrobot = model_free_acrobot_DDP(initial_state, final_state, MODEL_XML, alpha, horizon, state_dimemsion, control_dimension, Q, Q_final, R)

	# ---------------------------------------------------------------------------------------------------------
	# -----------------------------------------------Training---------------------------------------------------
	# ---------------------------------------------------------------------------------------------------------
	# Train the policy

	training_flag_on = True

	if training_flag_on:

		with open(path_to_data, 'w') as f:

			f.write("D2C training performed for a acrobot motion planning task:\n\n")

			f.write("System details : {}\n".format(os.uname().sysname + "--" + os.uname().nodename + "--" + os.uname().release + "--" + os.uname().version + "--" + os.uname().machine))
			f.write("-------------------------------------------------------------\n")

		time_1 = time.time()

		# # Run the DDP algorithm
		acrobot.iterate_ddp(n_iterations, finite_difference_gradients_flag=True)
		
		LQR_feedback_flag = False

		if LQR_feedback_flag:
			acrobot.feedback(W_x_LQR, W_u_LQR, W_x_LQR_f)

		time_2 = time.time()

		D2C_algorithm_run_time = time_2 - time_1

		print("D2C-2 algorithm run time taken: ", D2C_algorithm_run_time)
		
		# Save the episodic cost
		with open(training_cost_data_file, 'w') as f:
			for cost in acrobot.episodic_cost_history:
				f.write("%s\n" % cost)

		# # Test the obtained policy
		acrobot.save_policy(path_to_file)
		
		with open(path_to_data, 'a') as f:

				f.write("\nTotal time taken: {}\n".format(D2C_algorithm_run_time))
				f.write("------------------------------------------------------------------------------------------------------------------------------------\n")

		print(acrobot.X_p[-1])
		
		# Plot the episodic cost during the training
		acrobot.plot_episodic_cost_history(save_to_path=path_to_exp+"/episodic_cost_training.png")
	

	# ---------------------------------------------------------------------------------------------------------
	# -----------------------------------------------Testing---------------------------------------------------
	# ---------------------------------------------------------------------------------------------------------
	# Test the obtained policy

	test_flag_on = False

	if test_flag_on:

		f = open(path_to_model_free_DDP + path_to_exp + "/acrobot_testing_data.txt", "w")

		def frange(start, stop, step):
		    i = start
		    while i < stop:
		        yield i
		        i += step
		
		u_max = 5.38

		for i in frange(0.0, 1.02, 0.02):

		    episode_reward_n = 0
		    Var_n = 0
		    terminal_mse = 0
		    Var_terminal_mse = 0
		    n_samples = 1000

		    for j in range(n_samples):	

		        terminal_state = D2C_pendulum.test_episode(render=0, path=path_to_file, noise_stddev=i*u_max)
		        terminal_mse += np.linalg.norm(terminal_state, axis=0)
		        Var_terminal_mse += (np.linalg.norm(terminal_state, axis=0))**2

		    terminal_mse_avg = terminal_mse/n_samples
		    Var_terminal_mse_avg = (1/(n_samples*(n_samples-1)))*(n_samples*Var_terminal_mse - terminal_mse**2)

		    std_dev_mse = np.sqrt(Var_terminal_mse_avg)

		    f.write(str(i)+",\t"+str(terminal_mse_avg[0])+",\t"+str(std_dev_mse[0])+"\n")

		f.close()