ABP_inference/dataset_prep.py at main · npicchiarelli/ABP_inference · 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
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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
'''
Prepares the dataset for the model training.'''

import itertools
import pathlib
import pickle as pkl

import numpy as np
import pandas as pd

def data_prep(dir: str, datestamp: str, timestep: float, measevery: float, L:float, R: float, datafname: str):
    #Directory defining
    sim_folder = pathlib.Path(dir)
    datafolder = sim_folder / datestamp / "data"

    #Directory reading and storing in dataframes
    path_list = []
    for path in datafolder.iterdir():
        if path.is_dir():
            num = path.name[3:]
            df_path = path / f"{datestamp}_run{num}.txt"
            path_list.append(df_path)

    df_list = []
    for path in path_list:
        df_list.append(pd.read_csv(path))

    # Taking downsampling into account
    times = np.unique(df_list[0].Time)
    downsampling_factor = times[1] - times[0]
    timestep *= downsampling_factor*measevery

    #Velocity calculation and NaN dropping
    for df in df_list:
        df.sort_values(by=["N","Time"],inplace=True)
        df[["vx","vy"]] = df.groupby("N")[["xpos","ypos"]].diff()/timestep #velocity calculation
        #Accounting for PBC
        df.loc[df["vx"].abs() > (L-R)/timestep, 'vx'] -= np.sign(df[df["vx"].abs() > (L-R)/timestep].vx)*(L/timestep)
        df.loc[df["vy"].abs() > (L-R)/timestep, 'vy'] -= np.sign(df[df["vy"].abs() > (L-R)/timestep].vy)*(L/timestep)

        df[["ax","ay"]] = df.groupby("N")[["vx","vy"]].diff()/timestep #acceleration calculation
        df.dropna(inplace=True)
        df["Time"]=df["Time"]-(2*downsampling_factor)

    #Velocity, position, force, angle  reshaping
    num_exp = len(df_list)
    Np = df_list[0].N.max()
    num_steps = len(np.unique(df_list[0].Time))

    ppos = []
    vv = []
    angle_ = []
    force_ = []
    acc_ = []

    for df in df_list:
        Np = df.N.max()
        x = np.array(df.xpos).reshape(Np,-1)
        y = np.array(df.ypos).reshape(Np,-1)
        xy = np.stack([x,y],axis=2)
        ppos.append(xy)

        fx = np.array(df.fx).reshape(Np,-1)
        fy = np.array(df.fy).reshape(Np,-1)
        fxy = np.stack([fx,fy],axis=2)
        force_.append(fxy)

        vx = np.array(df.vx).reshape(Np,-1)
        vy = np.array(df.vy).reshape(Np,-1)
        vxy = np.stack([vx,vy],axis=2)
        vv.append(vxy)

        ax = np.array(df.ax).reshape(Np,-1)
        ay = np.array(df.ay).reshape(Np,-1)
        axy = np.stack([ax,ay],axis=2)
        acc_.append(axy)

        theta = np.array(df.orientation).reshape(Np,-1)
        angle_.append(theta)

    position = np.stack(ppos,axis=3)
    position = np.transpose(position,(3,1,0,2))

    force = np.stack(force_,axis=3)
    force = np.transpose(force,(3,1,0,2))

    vel = np.stack(vv,axis=3)
    vel = np.transpose(vel,(3,1,0,2))

    acc = np.stack(acc_,axis=3)
    acc = np.transpose(acc,(3,1,0,2))

    angle = np.stack(angle_, axis=2)[:,np.newaxis]
    angle = np.transpose(angle,(3,2,0,1))

    #Drag calculation
    gamma = 6*np.pi*R*1e-3
    gamma = np.repeat(gamma, Np)
    gamma = gamma[:,np.newaxis]
    gamma = np.tile(gamma,(num_exp,num_steps,1,1))

    #Edges
    edges = list(itertools.combinations(range(Np), 2))
    edges = np.array(edges)

    data = {
    "position": position,
    "velocity": vel,
    "acceleration": acc,
    "orientation":angle,
    "drag_coefficient": gamma,
    "node_force": force,
    "edge_list": edges,
    "dt": timestep
    }

    fdata = datafname+'.pkl'
    pathlib.Path(fdata).touch()

    with open(fdata, '+rb') as f:
        pkl.dump(data, f)

def data_prep_nosave(dir: str, datestamp: str, timestep: float, measevery: float, L:float, R: float):
    #Directory defining
    sim_folder = pathlib.Path(dir)
    datafolder = sim_folder / datestamp / "data"

    #Directory reading and storing in dataframes
    path_list = []
    for path in datafolder.iterdir():
        if path.is_dir():
            num = path.name[3:]
            df_path = path / f"{datestamp}_run{num}.txt"
            path_list.append(df_path)

    df_list = []
    for path in path_list:
        df_list.append(pd.read_csv(path))

    # Taking downsampling into account
    times = np.unique(df_list[0].Time)
    downsampling_factor = times[1] - times[0]
    timestep *= downsampling_factor*measevery

    #Velocity calculation and NaN dropping
    for df in df_list:
        df.sort_values(by=["N","Time"],inplace=True)
        df[["vx","vy"]] = df.groupby("N")[["xpos","ypos"]].diff()/timestep #velocity calculation
        df[["ax","ay"]] = df.groupby("N")[["vx","vy"]].diff()/timestep #acceleration calculation
        # Accounting for PBC
        df.loc[df["vx"].abs() > (L-R)/timestep, 'vx'] -= np.sign(df[df["vx"].abs() > (L-R)/timestep].vx)*(L/timestep)
        df.loc[df["vy"].abs() > (L-R)/timestep, 'vy'] -= np.sign(df[df["vy"].abs() > (L-R)/timestep].vy)*(L/timestep)
        df.dropna(inplace=True)
        df["Time"]=df["Time"]-(2*downsampling_factor)

    #Velocity, position, force, angle  reshaping
    num_exp = len(df_list)
    Np = df_list[0].N.max()
    num_steps = len(np.unique(df_list[0].Time))

    ppos = []
    vv = []
    angle_ = []
    force_ = []
    acc_ = []

    for df in df_list:
        Np = df.N.max()
        x = np.array(df.xpos).reshape(Np,-1)
        y = np.array(df.ypos).reshape(Np,-1)
        xy = np.stack([x,y],axis=2)
        ppos.append(xy)

        fx = np.array(df.fx).reshape(Np,-1)
        fy = np.array(df.fy).reshape(Np,-1)
        fxy = np.stack([fx,fy],axis=2)
        force_.append(fxy)

        vx = np.array(df.vx).reshape(Np,-1)
        vy = np.array(df.vy).reshape(Np,-1)
        vxy = np.stack([vx,vy],axis=2)
        vv.append(vxy)

        ax = np.array(df.ax).reshape(Np,-1)
        ay = np.array(df.ay).reshape(Np,-1)
        axy = np.stack([ax,ay],axis=2)
        acc_.append(axy)

        theta = np.array(df.orientation).reshape(Np,-1)
        angle_.append(theta)

    position = np.stack(ppos,axis=3)
    position = np.transpose(position,(3,1,0,2))

    force = np.stack(force_,axis=3)
    force = np.transpose(force,(3,1,0,2))

    vel = np.stack(vv,axis=3)
    vel = np.transpose(vel,(3,1,0,2))

    acc = np.stack(acc_,axis=3)
    acc = np.transpose(acc,(3,1,0,2))

    angle = np.stack(angle_, axis=2)[:,np.newaxis]
    angle = np.transpose(angle,(3,2,0,1))

    #Drag calculation
    gamma = 6*np.pi*R*1e-3
    gamma = np.repeat(gamma, Np)
    gamma = gamma[:,np.newaxis]
    gamma = np.tile(gamma,(num_exp,num_steps,1,1))

    #Edges
    edges = list(itertools.combinations(range(Np), 2))
    edges = np.array(edges)

    data = {
    "position": position,
    "velocity": vel,
    "acceleration": acc,
    "orientation":angle,
    "drag_coefficient": gamma,
    "node_force": force,
    "edge_list": edges,
    "dt": timestep
    }

    return data