Skip to content

W
wendelin
Commit 93dfd2fa authored by Léo-Paul Géneau's avatar Léo-Paul Géneau 👾
Browse files
Options

erp5_wendelin_drone: automatically fetch the amount of parameters

parent 477cf068
Expand all Show whitespace changes
Inline Side-by-side
Showing with 183 additions and 389 deletions
bt5/erp5_wendelin_drone/PathTemplateItem/portal_callables/DataAnalysisLine_convertDroneDataArrayToScoreArray.py
View file @ 93dfd2fa
......@@ -74,32 +74,6 @@ seen_sims = progress_indicator_sim.getStringOffsetIndex()
new_seen = ""
if seen_sims is None:
seen_sims = ""
score_dtypes = {'name': 'S256', 'Param1': 'f16', 'Param2': 'f16',
'distance_reciprocal': 'f8', 'ASML_reciprocal': 'f8',
'ground_speed_reciprocal': 'f8', 'climb_rate_reciprocal': 'f8',
'score_reciprocal': 'f16', 'score_cosine_row': 'f16',
'score_cosine_column': 'f16'}
_ = out_array_scores["Data Array"].initArray(shape=(0,), dtype=list(score_dtypes.items()))
plot_dtypes = {
'name': 'S256',
'Param1': 'f8',
'Param2': 'f8',
'timestamp': 'f8',
'distance_diff': 'f8',
'ASML_diff': 'f8',
'ground_speed_diff': 'f8',
'climb_rate_diff': 'f8',
'distance_reciprocal': 'f8',
'ASML_reciprocal': 'f8',
'ground_speed_reciprocal': 'f8',
'climb_rate_reciprocal': 'f8',
'score_reciprocal': 'f16',
'score_cosine_row': 'f16',
'score_cosine_column': 'f16'
}
_ = out_array_plot_data["Data Array"].initArray(shape=(0,), dtype=list(plot_dtypes.items()))
simulated_flight_list = []
simulated_flights_value_dict_list = []
......@@ -135,11 +109,11 @@ for name in not_seen_list:
break
try:
under_timestamp = simulated_flight[simulated_flight['timestamp (ms)'] - timestamp_diff < row['timestamp (ms)']].tail(1)
under_timestamp = simulated_flight[simulated_flight['timestamp (ms)'] - timestamp_diff < row['timestamp (ms)']].tail(1).to_dict('records')[0]
except IndexError:
continue
over_timestamp = simulated_flight[simulated_flight['timestamp (ms)'] - timestamp_diff >= row['timestamp (ms)']].head(1)
rate = (over_timestamp["timestamp (ms)"] - timestamp_diff - row["timestamp (ms)"])/(over_timestamp["timestamp (ms)"] - under_timestamp["timestamp (ms)"])
over_timestamp = simulated_flight[simulated_flight['timestamp (ms)'] - timestamp_diff >= row['timestamp (ms)']].head(1).to_dict('records')[0]
rate = (over_timestamp['timestamp (ms)'] - timestamp_diff - row['timestamp (ms)'])/(over_timestamp['timestamp (ms)'] - under_timestamp['timestamp (ms)'])
for criterion in distance_criterion_tuple:
tmp_sim[criterion].append(rated_value(over_timestamp[criterion], under_timestamp[criterion], rate))
......@@ -154,8 +128,9 @@ for name in not_seen_list:
),
tmp_sim['AMSL (m)'][-1],
tmp_sim['ground speed (m/s)'][-1],
tmp_real['climb rate (m/s)'][-1],
tmp_real['climb rate (m/s)'][-1]
)):
distance_list_tuple[index].append(value)
# If we have at least 100 entries (timestamps) analysed, add the filename to the list
......@@ -163,25 +138,19 @@ for name in not_seen_list:
if len(distance_list_tuple[0]) > 0:
# Add it to the dictionary
# A list of simulations, and their values over time
pattern = r'\((.*?)\)'
match = re.search(pattern, name)
if match:
values_str = match.group(1)
parameters = [float(val.strip()) for val in values_str.split(',')]
parameters = {k: float(v) for _, k, v in re.findall(r'[_-](([a-zA-Z_]+)_(\d+))', name)}
if parameters:
reciprocal_of_difference = [1/(1+mean_operation(list(map(abs, x)))) for x in distance_list_tuple[1:]]
score_reciprocal = sum(reciprocal_of_difference)
tmp_df_sim = pd.DataFrame(tmp_sim)
tmp_df_real = pd.DataFrame(tmp_real)
# Uncomment once you can actually use np..norm
score_cosine_row =mean_cosine_similarity_custom(tmp_df_real, tmp_df_sim)# cosine_similarity_rows(tmp_df_real, tmp_df_sim).mean()
score_cosine_row = mean_cosine_similarity_custom(tmp_df_real, tmp_df_sim)# cosine_similarity_rows(tmp_df_real, tmp_df_sim).mean()
score_cosine_column = mean_cosine_similarity_custom(tmp_df_real.T, tmp_df_sim.T)#cosine_similarity_rows(tmp_df_real.T, tmp_df_sim.T).mean()
simulated_flights_value_dict_list.append({
flights_value_dict = {
"name": name,
"Param1" : parameters[0],
"Param2": parameters[1],
"timestamp": distance_list_tuple[0],
"distance_diff": distance_list_tuple[1],
"ASML_diff" : distance_list_tuple[2],
......@@ -193,21 +162,24 @@ for name in not_seen_list:
"climb_rate_reciprocal" : reciprocal_of_difference[3],
"score_reciprocal": score_reciprocal/4,
"score_cosine_row" : score_cosine_row,
"score_cosine_column" : score_cosine_column
})
"score_cosine_column" : score_cosine_column,
}
selected_simulation_data_dict_list.append({
simulation_data_dict = {
"name": name,
"Param1" : parameters[0],
"Param2": parameters[1],
"distance_reciprocal": reciprocal_of_difference[0],
"ASML_reciprocal" : reciprocal_of_difference[1],
"ground_speed_reciprocal" : reciprocal_of_difference[2],
"climb_rate_reciprocal" : reciprocal_of_difference[3],
"score_reciprocal": score_reciprocal/4,
"score_cosine_row" : score_cosine_row,
"score_cosine_column" : score_cosine_column
})
}
for flight_parameter, value in parameters.items():
flights_value_dict[flight_parameter] = value
simulation_data_dict[flight_parameter] = value
simulated_flights_value_dict_list.append(flights_value_dict)
selected_simulation_data_dict_list.append(simulation_data_dict)
combined_data = pd.DataFrame() # Initialize an empty DataFrame
......@@ -217,26 +189,7 @@ for data_dict in simulated_flights_value_dict_list:
tmp = pd.DataFrame(data_dict) # Wrap data_dict in a list and create a DataFrame
combined_data = pd.concat([combined_data, tmp], ignore_index=True)
plot_dtypes = {
'name': 'S256',
'Param1': 'float64',
'Param2': 'float64',
'timestamp': 'float64',
'distance_diff': 'float64',
'ASML_diff': 'float64',
'ground_speed_diff': 'float64',
'climb_rate_diff': 'float64',
'distance_reciprocal': 'float64',
'ASML_reciprocal': 'float64',
'ground_speed_reciprocal': 'float64',
'climb_rate_reciprocal': 'float64',
'score_reciprocal': 'float64',
'score_cosine_row': 'float64',
'score_cosine_column': 'float64',
}
plot_dtypes = {k: 'float64' if k != 'name' else 'S256' for k in simulated_flights_value_dict_list[0]}
plots_df = combined_data
if plots_df.empty:
......@@ -261,11 +214,7 @@ for data_dict in selected_simulation_data_dict_list:
score_dtypes = {'name': 'S256', 'Param1': 'float64', 'Param2': 'float64',
'distance_reciprocal': 'float64', 'ASML_reciprocal': 'float64',
'ground_speed_reciprocal': 'float64', 'climb_rate_reciprocal': 'float64',
'score_reciprocal': 'float64', 'score_cosine_row': 'float64',
'score_cosine_column': 'float64'}
score_dtypes = {k: 'float64' if k != 'name' else 'S256' for k in selected_simulation_data_dict_list[0]}
scores_df = pd.DataFrame(combined_data)
scores_df = scores_df.astype(score_dtypes)
###############################
......@@ -273,6 +222,9 @@ scores_df = scores_df.astype(score_dtypes)
zbigarray_scores = out_array_scores["Data Array"].getArray()
if zbigarray_scores is None:
out_array_scores["Data Array"].initArray(shape=(0,), dtype=list(score_dtypes.items()))
out_array_plot_data["Data Array"].initArray(shape=(0,), dtype=list(plot_dtypes.items()))
score_names = list(out_array_scores["Data Array"])
......@@ -281,7 +233,7 @@ score_names = list(out_array_scores["Data Array"])
scores_ndarray = scores_df.to_records(index = False)
scores_ndarray = scores_ndarray.astype([("name", "S256"),("Param1", "f8"), ("Param2", "f8"), ("distance_reciprocal", "f8"), ("ASML_reciprocal", "f8"), ("ground_speed_reciprocal", "f8"), ('climb_rate_reciprocal', 'f8'), ("score_reciprocal", "f8"), ("score_cosine_column", "f8"), ("score_cosine_row", "f8")])
scores_ndarray = scores_ndarray.astype([(k, 'f8' if k != 'name' else 'S256') for k in score_dtypes])
if zbigarray_scores is None:
......@@ -330,7 +282,7 @@ zbigarray_plots = out_array_plot_data["Data Array"].getArray()
plots_ndarray = plots_df.to_records(index = False)
plots_ndarray = plots_ndarray.astype([('name', 'S256'), ('Param1', 'f8'), ('Param2', 'f8'), ('timestamp', 'f8'), ('distance_diff', 'f8'), ('ASML_diff', 'f8'), ('ground_speed_diff', 'f8'), ('climb_rate_diff', 'f8'), ('distance_reciprocal', 'f8'), ('ASML_reciprocal', 'f8'), ('ground_speed_reciprocal', 'f8'), ('climb_rate_reciprocal', 'f8'), ('score_reciprocal', 'f8'), ('score_cosine_row', 'f8'), ('score_cosine_column', 'f8')])
plots_ndarray = plots_ndarray.astype([(k, 'f8' if k != 'name' else 'S256') for k in plot_dtypes])
if zbigarray_plots is None:
zbigarray_plots = out_array_plot_data["Data Array"].initArray(shape=(0,), dtype=plots_ndarray.dtype.fields)
......
bt5/erp5_wendelin_drone/PathTemplateItem/portal_callables/DataAnalysisLine_recalculateScore.py
View file @ 93dfd2fa
......@@ -8,148 +8,64 @@ import re
# Now, to save space, we just overwrite the old array with the new one, that contains the new scores
# This new array will give us the final overview of the ranking of the parameters. This can be used by our genetic algorithm to decide when we can stop and which parameters worked the best.
score_dtypes = {'name': 'S256', 'Param1': 'f8', 'Param2': 'f8',
'distance_reciprocal': 'f8', 'ASML_reciprocal': 'f8',
'ground_speed_reciprocal': 'f8', 'climb_rate_reciprocal': 'f8',
'score_reciprocal': 'f8', 'score_cosine_row': 'f8',
'score_cosine_column': 'f8'}
new_score_dtypes= {'name': 'S256', 'Param1': 'f8', 'Param2': 'f8',
'distance_reciprocal': 'f8', 'ASML_reciprocal': 'f8',
'ground_speed_reciprocal': 'f8', 'climb_rate_reciprocal': 'f8',
'score_reciprocal': 'f8', 'score_cosine_row': 'f8',
'score_cosine_column': 'f8',
'iteration': 'i8'}
plot_dtypes = {
'name': 'S256',
'Param1': 'f8',
'Param2': 'f8',
'timestamp': 'f8',
'distance_diff': 'f8',
'ASML_diff': 'f8',
'ground_speed_diff': 'f8',
'climb_rate_diff': 'f8',
'distance_reciprocal': 'f8',
'ASML_reciprocal': 'f8',
'ground_speed_reciprocal': 'f8',
'climb_rate_reciprocal': 'f8',
'score_reciprocal': 'f8',
'score_cosine_row': 'f8',
'score_cosine_column': 'f8',
'iteration': 'i8'
}
new_plot_dtypes = {
'Param1': 'f8',
'Param2': 'f8',
'timestamp': 'f8',
'distance_diff': 'f8',
'ASML_diff': 'f8',
'ground_speed_diff': 'f8',
'climb_rate_diff': 'f8',
'distance_reciprocal': 'f8',
'ASML_reciprocal': 'f8',
'ground_speed_reciprocal': 'f8',
'climb_rate_reciprocal': 'f8',
'score_reciprocal': 'f8',
'score_cosine_row': 'f8',
'score_cosine_column': 'f8',
'iteration': 'i8'
}
score_array = input_array_scores["Data Array"]
new_score_array = out_array_scores["Data Array"]
plot_array = input_array_plot["Data Array"]
new_plot_array = out_array_plots["Data Array"]
# Should only look at the newest few
score_array = input_array_scores['Data Array']
score_nparray = score_array.getArray()
if score_nparray is None:
score_nparray = out_array_scores["Data Array"].initArray(shape=(0,), dtype=list(new_score_dtypes.items()))
return
plot_array = input_array_plot['Data Array']
plot_nparray = plot_array.getArray()
if plot_nparray is None:
plot_nparray = out_array_plots["Data Array"].initArray(shape=(0,), dtype=list(new_plot_dtypes.items()))
return
new_score_dtypes = [(k, v[0]) for k, v in dict(score_nparray.dtype.fields).items()]
new_score_dtypes.append(('iteration', 'i8'))
new_score_array = out_array_scores['Data Array']
new_score_nparray = new_score_array.getArray()
if new_score_nparray is None:
new_score_array.initArray(shape=(0,), dtype=new_score_dtypes)
new_plot_dtypes = [(k, v[0]) for k, v in dict(plot_nparray.dtype.fields).items()]
new_plot_dtypes.append(('iteration', 'i8'))
new_plot_array = out_array_plots['Data Array']
new_plot_nparray = new_plot_array.getArray()
if new_plot_nparray is None:
new_plot_array.initArray(shape=(0,), dtype=new_plot_dtypes)
old_score_df = pd.DataFrame.from_records(score_nparray[:].copy())
old_plot_df = pd.DataFrame.from_records(plot_nparray[:].copy())
progress_indicator = input_array_scores["Progress Indicator"]
progress_indicator = input_array_scores['Progress Indicator']
seen_sims = progress_indicator.getStringOffsetIndex()
if seen_sims is None:
seen_sims = ""
sim_flight_names = list(old_score_df["name"])
sim_flight_names = list(old_score_df['name'])
# We will only continue if there is new data available.
if len([x for x in sim_flight_names if x not in seen_sims]) == 0:
return
new_score_nparray = new_score_array.getArray()
if new_score_nparray is None:
new_score_nparray = out_array_scores["Data Array"].initArray(shape=(0,), dtype=list(new_score_dtypes.items()))
new_plot_nparray = new_plot_array.getArray()
if new_plot_nparray is None:
new_plot_nparray = out_array_plots["Data Array"].initArray(shape=(0,), dtype=list(new_plot_dtypes.items()))
new_score_df = pd.DataFrame.from_records(new_score_nparray[:].copy())
new_plot_df = pd.DataFrame.from_records(new_plot_nparray[:].copy())
new_score_iteration = new_score_df["iteration"].max()
new_score_iteration = new_score_df['iteration'].max()
if math.isnan(new_score_iteration):
new_score_iteration = 0
new_score_df = new_score_df.drop(["iteration"],axis=1)
new_score_df = new_score_df.drop(['iteration'], axis=1)
new_plot_iteration = new_plot_df["iteration"].max()
if math.isnan(new_plot_iteration):
new_plot_iteration = 0
#new_plot_df = new_plot_df.drop(columns=["iteration"])
answer_scores = pd.concat([old_score_df, new_score_df]).drop_duplicates(subset=['Param1','Param2'], keep='last')
answer_plots = pd.concat([old_plot_df, new_plot_df]).drop_duplicates(subset=['Param1','Param2'], keep='last')
answer_scores["score_reciprocal"] = answer_scores["score_reciprocal"].astype('float64')
answer_scores = answer_scores.nlargest(5, "score_reciprocal")
answer_plots = old_plot_df[old_plot_df["name"].isin(list(answer_scores["name"]))]
param_subset = [k for k in old_score_df.dtypes.keys().values if not 'reciprocal' in k and not 'score' in k and k != 'name']
answer_scores = pd.concat([old_score_df, new_score_df]).drop_duplicates(subset=param_subset, keep='last')
answer_plots = pd.concat([old_plot_df, new_plot_df]).drop_duplicates(subset=param_subset, keep='last')
answer_scores['score_reciprocal'] = answer_scores['score_reciprocal'].astype('float64')
answer_scores = answer_scores.nlargest(5, 'score_reciprocal')
answer_plots = old_plot_df[old_plot_df['name'].isin(list(answer_scores['name']))]
new_score_iteration = new_score_iteration + 1
new_plot_iteration = new_plot_iteration + 1
......@@ -158,24 +74,14 @@ answer_scores["iteration"] = new_score_iteration
answer_plots["iteration"] = new_plot_iteration
# We will remove all the data from the data arrays before we append the new data. Essentially we will be left with only the best few iterations
new_score_nparray = out_array_scores["Data Array"].initArray(shape=(0,), dtype=list(new_score_dtypes.items()))
new_plot_nparray = out_array_plots["Data Array"].initArray(shape=(0,), dtype=list(new_plot_dtypes.items()))
new_score_nparray = new_score_array.initArray(shape=(0,), dtype=new_score_dtypes)
new_plot_nparray = new_plot_array.initArray(shape=(0,), dtype=new_plot_dtypes)
new_score_nparray.append(answer_scores.to_records(index = False))
# Group the DataFrame by the 'name' column
grouped = answer_plots.groupby(['Param1', 'Param2'])
grouped = answer_plots.groupby(param_subset)
# Initialize a list to store the resulting DataFrames
resulting_dfs = []
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7