marl-demandresponse-original/plotting.py at main · maivincent/marl-demandresponse-original · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
#%% Imports

import matplotlib.pyplot as plt
import matplotlib.colors as clr
import numpy as np
import os
import random
import uuid
import wandb

from utils import normStateDict

#%% Functions

def plot_env_test(env, action_type='off', n_steps=1000):
    assert action_type in ['off', 'on', 'random'], 'Action types available: off/on/random'
    action_types = {'on': 1, 'off': 0, 'random': 0}

    # Reset environment
    obs_dict = env.reset()

    # Initialize arrays
    reward = np.empty(n_steps)
    hvac = np.empty(n_steps)
    temp = np.empty(n_steps)

    # Act on environment and save reward, hvac status and temperature
    for t in range(n_steps):
        if action_type == 'random':
            action = {"0_1": random.randint(0,1)}
        else:
            action = {"0_1": action_types[action_type]}
        next_obs_dict, rewards_dict, dones_dict, info_dict = env.step(action)

        # Save data in arrays
        reward[t] = rewards_dict["0_1"]
        hvac[t] = next_obs_dict["0_1"]["hvac_turned_on"]
        temp[t] = next_obs_dict["0_1"]["house_temp"]

    plt.scatter(np.arange(len(hvac)), hvac, s=1, marker='.', c='orange')
    plt.plot(reward)
    plt.title('HVAC state vs. Reward')
    plt.show()
    plt.plot(temp)
    plt.title('Temperature')
    plt.show()

def plot_agent_test(env, agent, config_dict, n_steps=1000):
    # Reset environment
    obs_dict = env.reset()
    cumul_avg_reward = 0

    # Initialize arrays
    reward = np.empty(n_steps)
    hvac = np.empty(n_steps)
    actions = np.empty(n_steps)
    temp = np.empty(n_steps)

    # Act on environment and save reward, hvac status and temperature
    for t in range(n_steps):
        action = {"0_1": agent.select_action(normStateDict(obs_dict["0_1"], config_dict))}
        next_obs_dict, rewards_dict, dones_dict, info_dict = env.step(action)

        # Save data in arrays
        actions[t] = action["0_1"]
        reward[t] = rewards_dict["0_1"]
        hvac[t] = next_obs_dict["0_1"]["hvac_turned_on"]
        temp[t] = next_obs_dict["0_1"]["house_temp"]

        cumul_avg_reward += rewards_dict["0_1"] / env.nb_agents

        obs_dict = next_obs_dict

    print(cumul_avg_reward/n_steps)

    fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(15,5))
    ax1.plot(actions)
    ax1.plot(hvac)
    ax1.title.set_text('HVAC state vs. Agent action')
    ax2.plot(reward)
    ax2.title.set_text("Reward")
    ax3.plot(temp)
    ax3.title.set_text('Temperature')
    plt.show()

#%%

def colorPlotTestAgentHouseTemp(prob_on_per_training_on, prob_on_per_training_off, low_temp, high_temp, time_steps_test_log, log_wandb):
    '''
    Makes a color plot of the probability of the agent to turn on given indoors temperature, with the training
    '''
    prob_on_per_training_on = prob_on_per_training_on[1:]
    prob_on_per_training_off = prob_on_per_training_off[1:]

    fig, axes = plt.subplots(nrows=2, ncols=1, figsize=(8,8.5), dpi=80)
    print(axes)

    normalizer = clr.Normalize(vmin=0, vmax=1)
    map0 = axes[0].imshow(np.transpose(prob_on_per_training_on), extent=[0, np.size(prob_on_per_training_on, 1)*time_steps_test_log, high_temp, low_temp], norm=normalizer)
    map1 = axes[1].imshow(np.transpose(prob_on_per_training_off), extent=[0, np.size(prob_on_per_training_off, 1)*time_steps_test_log, high_temp, low_temp], norm=normalizer)
    #axes[0] = plt.gca()
    axes[0].invert_yaxis()
    axes[1].invert_yaxis()

    forceAspect(axes[0], aspect=2.0)
    forceAspect(axes[1], aspect=2.0)

    axes[0].set_xlabel("Training time steps")
    axes[1].set_xlabel("Training time steps")
    axes[0].set_ylabel("Indoors temperature")
    axes[1].set_ylabel("Indoors temperature")
    axes[0].set_title("Power: ON")
    axes[1].set_title("Power: OFF")

    cb = fig.colorbar(map0, ax=axes[:], shrink=0.6)

    if log_wandb:
        name = uuid.uuid1().hex + "probTestAgent.png"
        plt.savefig(name)
        wandb.log(
            {"Probability of agent vs Indoor temperature vs Episode ": wandb.Image(name)})
        os.remove(name)

    else:
        plt.show()
    return 0

def forceAspect(ax, aspect):
    im = ax.get_images()
    extent = im[0].get_extent()
    ax.set_aspect(abs((extent[1]-extent[0])/(extent[3]-extent[2]))/aspect)