1
0
mirror of https://github.com/osmarks/random-stuff synced 2024-11-08 13:39:53 +00:00
random-stuff/pid-controller.py
2023-02-05 21:59:30 +00:00

43 lines
1.8 KiB
Python

import time
import collections
PIDState = collections.namedtuple("PIDState", ["Kp", "Ki", "Kd", "last_time", "integral", "last_error"])
def init(Kp, Ki, Kd): return PIDState(Kp, Ki, Kd, None, 0, None)
def step(state, error, deriv, ts=None):
ts = ts if ts is not None else time.time()
integral = state.integral
if state.last_time:
tdiff = ts - state.last_time
integral += 0.5 * (state.last_error + error) * tdiff # approximate actually integrating using a trapzeium
output = state.Kp * error + state.Ki * integral + state.Kd * deriv
return PIDState(Kp=state.Kp, Ki=state.Ki, Kd=state.Kd, last_time=ts, integral=integral, last_error=error), output
if __name__ == "__main__":
import matplotlib.pyplot as plt, numpy as np
def extract_series(l, ix): return list(map(lambda x: x[ix], l))
values = [(10, -10, 0, 0, 0)]
state = init(10, 4, -0.3)
setpoint = -5
max_time = 2
timestep = 0.05
times = np.arange(0, max_time, timestep)
for t in times:
current_pv = values[-1][0]
error = setpoint - current_pv
deriv = (error - (state.last_error or 0)) / timestep
state, output = step(state, error, deriv, ts=t)
output = max(min(output, 10), -10)
print(output, current_pv, error)
new_pv = current_pv + (output + 1) * 0.05
values.append((new_pv, error, output, state.integral, deriv))
#print(values)
values = values[1:]
plt.axis([0, max_time, -10, 10])
plt.plot(times, extract_series(values, 0), label="PV")
plt.plot(times, extract_series(values, 1), label="error")
plt.plot(times, extract_series(values, 2), label="output")
plt.plot(times, extract_series(values, 3), label="integ")
plt.plot(times, extract_series(values, 4), label="deriv")
plt.legend()
plt.show()