openpilot is an open source driver assistance system. openpilot performs the functions of Automated Lane Centering and Adaptive Cruise Control for over 200 supported car makes and models.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

122 lines
3.1 KiB

#!/usr/bin/env python3
'''
printing the gap between interrupts in a histogram to check if the
frequency is what we expect, the bmx is not interrupt driven for as we
get interrupts in a 2kHz rate.
'''
import argparse
import sys
from collections import defaultdict
from tools.lib.logreader import LogReader
from tools.lib.route import Route
import matplotlib.pyplot as plt
SRC_BMX = "bmx055"
SRC_LSM = "lsm6ds3"
def parseEvents(log_reader):
bmx_data = defaultdict(list)
lsm_data = defaultdict(list)
for m in log_reader:
# only sensorEvents
if m.which() != 'sensorEvents':
continue
for se in m.sensorEvents:
# convert data to dictionary
d = se.to_dict()
if d["timestamp"] == 0:
continue # empty event?
if d["source"] == SRC_BMX and "acceleration" in d:
bmx_data["accel"].append(d["timestamp"] / 1e9)
if d["source"] == SRC_BMX and "gyroUncalibrated" in d:
bmx_data["gyro"].append(d["timestamp"] / 1e9)
if d["source"] == SRC_LSM and "acceleration" in d:
lsm_data["accel"].append(d["timestamp"] / 1e9)
if d["source"] == SRC_LSM and "gyroUncalibrated" in d:
lsm_data["gyro"].append(d["timestamp"] / 1e9)
return bmx_data, lsm_data
def cleanData(data):
if len(data) == 0:
return [], []
data.sort()
prev = data[0]
diffs = []
for v in data[1:]:
diffs.append(v - prev)
prev = v
return data, diffs
def logAvgValues(data, sensor):
if len(data) == 0:
print(f"{sensor}: no data to average")
return
avg = sum(data) / len(data)
hz = 1 / avg
print(f"{sensor}: data_points: {len(data)} avg [ns]: {avg} avg [Hz]: {hz}")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("route", type=str, help="route name")
parser.add_argument("segment", type=int, help="segment number")
args = parser.parse_args()
r = Route(args.route)
logs = r.log_paths()
if len(logs) == 0:
print("NO data routes")
sys.exit(0)
if args.segment >= len(logs):
print(f"RouteID: {args.segment} out of range, max: {len(logs) -1}")
sys.exit(0)
lr = LogReader(logs[args.segment])
bmx_data, lsm_data = parseEvents(lr)
# sort bmx accel data, and then cal all the diffs, and to a histogram of those
bmx_accel, bmx_accel_diffs = cleanData(bmx_data["accel"])
bmx_gyro, bmx_gyro_diffs = cleanData(bmx_data["gyro"])
lsm_accel, lsm_accel_diffs = cleanData(lsm_data["accel"])
lsm_gyro, lsm_gyro_diffs = cleanData(lsm_data["gyro"])
# get out the averages
logAvgValues(bmx_accel_diffs, "bmx accel")
logAvgValues(bmx_gyro_diffs, "bmx gyro ")
logAvgValues(lsm_accel_diffs, "lsm accel")
logAvgValues(lsm_gyro_diffs, "lsm gyro ")
fig, axs = plt.subplots(1, 2, tight_layout=True)
axs[0].hist(bmx_accel_diffs, bins=50)
axs[0].set_title("bmx_accel")
axs[1].hist(bmx_gyro_diffs, bins=50)
axs[1].set_title("bmx_gyro")
figl, axsl = plt.subplots(1, 2, tight_layout=True)
axsl[0].hist(lsm_accel_diffs, bins=50)
axsl[0].set_title("lsm_accel")
axsl[1].hist(lsm_gyro_diffs, bins=50)
axsl[1].set_title("lsm_gyro")
print("check plot...")
plt.show()