forked from yfzhang/vehicle-motion-forecasting
-
Notifications
You must be signed in to change notification settings - Fork 0
/
maxent_nonlinear_offroad_rank.py
169 lines (146 loc) · 9.13 KB
/
maxent_nonlinear_offroad_rank.py
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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import numpy as np
np.set_printoptions(threshold=np.inf) # print the full numpy array
import warnings
warnings.filterwarnings('ignore')
from torch.autograd import Variable
import torch
from multiprocessing import Pool
from viz import overlay, feat2rgb
def overlay_traj_to_map(traj1, traj2, feat, value1=5.0, value2=10.0):
overlay_map = feat.copy()
for i, p in enumerate(traj1):
overlay_map[int(p[0]), int(p[1])] = value1
for i, p in enumerate(traj2):
overlay_map[int(p[0]), int(p[1])] = value2
return overlay_map
def visualize_batch(past_traj, future_traj, feat, r_var, values, svf_diff_var, optimal_traj_list, step, vis, grid_size, train=True):
mode = 'train' if train else 'test'
n_batch = past_traj.shape[0]
step = step*n_batch
for i in range(n_batch):
future_traj_sample = future_traj[i].numpy() # choose one sample from the batch
future_traj_sample = future_traj_sample[~np.isnan(future_traj_sample).any(axis=1)] # remove appended NAN rows
future_traj_sample = future_traj_sample.astype(np.int64)
past_traj_sample = past_traj[i].numpy() # choose one sample from the batch
past_traj_sample = past_traj_sample[~np.isnan(past_traj_sample).any(axis=1)] # remove appended NAN rows
past_traj_sample = past_traj_sample.astype(np.int64)
optimal_traj_sample = optimal_traj_list[i] # choose one sample from the batch
overlay_map = feat[i, 0, :, :].float().view(grid_size, -1).numpy() # (grid_size, grid_size)
past = np.min(feat[i, 0, :, :].numpy())
future = np.max(feat[i, 0, :, :].numpy())
overlay_map = overlay_traj_to_map(past_traj_sample, optimal_traj_sample, overlay_map, past, future)
vis.heatmap(X=np.flip(overlay_map, 0), opts=dict(colormap='Electric', title='{}, step {}, height'.format(mode, step+i)))
# save to file
np.savetxt('height_map/height_map_{step:02d}.txt'.format(step=step+i), X=feat[i, 0, :, :], fmt='%.2f')
with open('optimal_traj/optimal_traj_{step:02d}.txt'.format(step=step+i), 'w') as f:
for index in optimal_traj_sample:
f.writelines('{} {}\n'.format(index[0], index[1]))
overlay_map = overlay(feat2rgb(feat[i].numpy()), future_traj_sample, past_traj_sample)
vis.image(np.transpose(overlay_map, (2, 0, 1)), opts=dict(title='{} rgb'.format(step+i)))
vis.heatmap(X=np.flip(r_var.data[i].view(grid_size, -1), 0),
opts=dict(colormap='Electric', title='{}, step {}, rewards'.format(mode, step+i)))
vis.heatmap(X=np.flip(svf_diff_var.data[i].view(grid_size, -1), 0),
opts=dict(colormap='Electric', title='{}, step {}, SVF_diff'.format(mode, step+i)))
def visualize(past_traj, future_traj, feat, r_var, values, svf_diff_var, step, vis, grid_size, train=True):
mode = 'train' if train else 'test'
future_traj_sample = future_traj[0].numpy() # choose one sample from the batch
future_traj_sample = future_traj_sample[~np.isnan(future_traj_sample).any(axis=1)] # remove appended NAN rows
future_traj_sample = future_traj_sample.astype(np.int64)
past_traj_sample = past_traj[0].numpy() # choose one sample from the batch
past_traj_sample = past_traj_sample[~np.isnan(past_traj_sample).any(axis=1)] # remove appended NAN rows
past_traj_sample = past_traj_sample.astype(np.int64)
vis.heatmap(X=feat[0, 0, :, :].float().view(grid_size, -1),
opts=dict(colormap='Electric', title='{}, step {} height max'.format(mode, step)))
overlay_map = feat[0, 1, :, :].float().view(grid_size, -1).numpy() # (grid_size, grid_size)
overlay_map = overlay_traj_to_map(past_traj_sample, future_traj_sample, overlay_map)
vis.heatmap(X=overlay_map, opts=dict(colormap='Electric', title='{}, step {} height var'.format(mode, step)))
vis.heatmap(X=feat[0, 3, :, :].float().view(grid_size, -1),
opts=dict(colormap='Electric', title='{}, step {} green'.format(mode, step)))
vis.heatmap(X=r_var.data[0].view(grid_size, -1),
opts=dict(colormap='Greys', title='{}, step {}, rewards'.format(mode, step)))
vis.heatmap(X=values[0].reshape(grid_size, -1),
opts=dict(colormap='Greys', title='{}, step {}, value'.format(mode, step)))
vis.heatmap(X=svf_diff_var.data[0].view(grid_size, -1),
opts=dict(colormap='Greys', title='{}, step {}, SVF_diff'.format(mode, step)))
# for name, param in net.named_parameters():
# if name.endswith('weight'):
# vis.histogram(param.data.view(-1), opts=dict(numbins=20)) # weights
# vis.histogram(param.grad.data.view(-1), opts=dict(numbins=20)) # grads
def rl(future_traj_sample, r_sample, model, grid_size):
svf_demo_sample = model.find_demo_svf(future_traj_sample)
values_sample = model.find_optimal_value(r_sample, 0.1)
policy = model.find_stochastic_policy(values_sample, r_sample)
svf_sample = model.find_svf(future_traj_sample, policy)
svf_diff_sample = svf_demo_sample - svf_sample
# (1, n_feature, grid_size, grid_size)
svf_diff_sample = svf_diff_sample.reshape(1, 1, grid_size, grid_size)
svf_diff_var_sample = Variable(torch.from_numpy(svf_diff_sample).float(), requires_grad=False)
nll_sample = model.compute_nll(policy, future_traj_sample)
print(nll_sample)
return nll_sample, svf_diff_var_sample, values_sample
def rl_rank(future_traj_sample, past_traj_sample, r_sample, model, grid_size):
svf_demo_sample = model.find_demo_svf(future_traj_sample)
values_sample = model.find_optimal_value(r_sample, 0.1)
policy = model.find_stochastic_policy(values_sample, r_sample)
svf_sample = model.find_svf(future_traj_sample, policy)
svf_diff_sample = svf_demo_sample - svf_sample
# (1, n_feature, grid_size, grid_size)
svf_diff_sample = svf_diff_sample.reshape(1, 1, grid_size, grid_size)
svf_diff_var_sample = Variable(torch.from_numpy(svf_diff_sample).float(), requires_grad=False)
nll_sample = model.compute_nll(policy, future_traj_sample)
print(nll_sample)
past_return_sample = model.compute_return(r_sample, past_traj_sample) # compute return
past_return_sample = np.array([past_return_sample])
past_return_var_sample = Variable(torch.from_numpy(past_return_sample).float())
return nll_sample, svf_diff_var_sample, values_sample, past_return_var_sample
def pred(feat, robot_state_feat, future_traj, net, n_states, model, grid_size):
n_sample = feat.shape[0]
feat = feat.float()
feat_var = Variable(feat)
robot_state_feat = robot_state_feat.float()
robot_state_feat_var = Variable(robot_state_feat)
r_var = net(feat_var, robot_state_feat_var)
result = []
pool = Pool(processes=n_sample)
for i in range(n_sample):
r_sample = r_var[i].data.numpy().squeeze().reshape(n_states)
future_traj_sample = future_traj[i].numpy() # choose one sample from the batch
future_traj_sample = future_traj_sample[~np.isnan(future_traj_sample).any(axis=1)] # remove appended NAN rows
future_traj_sample = future_traj_sample.astype(np.int64)
result.append(pool.apply_async(rl, args=(future_traj_sample, r_sample, model, grid_size)))
pool.close()
pool.join()
# extract result and stack svf_diff
nll_list = [result[i].get()[0] for i in range(n_sample)]
svf_diff_var_list = [result[i].get()[1] for i in range(n_sample)]
values_list = [result[i].get()[2] for i in range(n_sample)]
svf_diff_var = torch.cat(svf_diff_var_list, dim=0)
return nll_list, r_var, svf_diff_var, values_list
def pred_rank(feat, robot_state_feat, future_traj, past_traj, net, n_states, model, grid_size):
n_sample = feat.shape[0]
feat = feat.float()
feat_var = Variable(feat)
robot_state_feat = robot_state_feat.float()
robot_state_feat_var = Variable(robot_state_feat)
r_var = net(feat_var, robot_state_feat_var)
result = []
pool = Pool(processes=n_sample)
for i in range(n_sample):
r_sample = r_var[i].data.numpy().squeeze().reshape(n_states)
future_traj_sample = future_traj[i].numpy() # choose one sample from the batch
future_traj_sample = future_traj_sample[~np.isnan(future_traj_sample).any(axis=1)] # remove appended NAN rows
future_traj_sample = future_traj_sample.astype(np.int64)
past_traj_sample = past_traj[i].numpy() # choose one sample from the batch
past_traj_sample = past_traj_sample[~np.isnan(past_traj_sample).any(axis=1)] # remove appended NAN rows
past_traj_sample = past_traj_sample.astype(np.int64)
result.append(pool.apply_async(rl_rank, args=(future_traj_sample, past_traj_sample, r_sample, model, grid_size)))
pool.close()
pool.join()
# extract result and stack svf_diff
nll_list = [result[i].get()[0] for i in range(n_sample)]
svf_diff_var_list = [result[i].get()[1] for i in range(n_sample)]
values_list = [result[i].get()[2] for i in range(n_sample)]
past_return_var_list = [result[i].get()[3] for i in range(n_sample)]
svf_diff_var = torch.cat(svf_diff_var_list, dim=0)
past_return_var = torch.cat(past_return_var_list, dim=0)
return nll_list, r_var, svf_diff_var, values_list, past_return_var