Files
onepilot/selfdrive/controls/lib/latcontrol_torque.py
T
firestar5683 d0e1db6766 StarPilot
2026-03-22 03:15:05 -05:00

207 lines
10 KiB
Python

import math
import numpy as np
from collections import deque
from cereal import log
from opendbc.car.gm.values import CAR as GM_CAR
from opendbc.car.lateral import get_friction
from openpilot.common.constants import ACCELERATION_DUE_TO_GRAVITY, CV
from openpilot.common.filter_simple import FirstOrderFilter
from openpilot.common.pid import PIDController
from openpilot.selfdrive.controls.lib.drive_helpers import MIN_SPEED
from openpilot.selfdrive.controls.lib.latcontrol import LatControl
# At higher speeds (25+mph) we can assume:
# Lateral acceleration achieved by a specific car correlates to
# torque applied to the steering rack. It does not correlate to
# wheel slip, or to speed.
# This controller applies torque to achieve desired lateral
# accelerations. To compensate for the low speed effects the
# proportional gain is increased at low speeds by the PID controller.
# Additionally, there is friction in the steering wheel that needs
# to be overcome to move it at all, this is compensated for too.
KP = 0.7
KI = 0.35
INTERP_SPEEDS = [1, 1.5, 2.0, 3.0, 5, 7.5, 10, 15, 30]
KP_INTERP = [250, 120, 65, 30, 11.5, 5.5, 3.5, 2.0, KP]
LOW_SPEED_X = [0, 10, 20, 30]
LOW_SPEED_Y = [12, 10.5, 8, 5]
MAX_LAT_JERK_UP = 2.5 # m/s^3
LP_FILTER_CUTOFF_HZ = 1.2
JERK_LOOKAHEAD_SECONDS = 0.19
JERK_GAIN = 0.22
LAT_ACCEL_REQUEST_BUFFER_SECONDS = 1.0
VERSION = 2
DEBUG_TORQUE_TUNE = False
FF_SCALE_BLEND_LAT_ACCEL = 0.05
DEADZONE_BOOST_LAT_ACCEL = 0.15
UNWIND_D_DES_THRESHOLD = -1.0
UNWIND_LAT_ACCEL_NEAR_ZERO = 0.3
MIN_LATERAL_CONTROL_SPEED = 0.3
BOLT_2022_2023_CARS = (
GM_CAR.CHEVROLET_BOLT_ACC_2022_2023,
GM_CAR.CHEVROLET_BOLT_ACC_2022_2023_PEDAL,
GM_CAR.CHEVROLET_BOLT_CC_2022_2023,
)
BOLT_2018_2021_CARS = (
GM_CAR.CHEVROLET_BOLT_CC_2018_2021,
)
BOLT_2017_CARS = (
GM_CAR.CHEVROLET_BOLT_CC_2017,
)
BOLT_CARS = BOLT_2022_2023_CARS + BOLT_2018_2021_CARS + BOLT_2017_CARS
def get_friction_threshold(v_ego: float) -> float:
# Keep StarPilot's speed-scaled friction threshold behavior.
return float(np.interp(v_ego, [1 * CV.MPH_TO_MS, 20 * CV.MPH_TO_MS, 75 * CV.MPH_TO_MS], [0.16, 0.19, 0.27]))
class LatControlTorque(LatControl):
def __init__(self, CP, CI, dt):
super().__init__(CP, CI, dt)
self.torque_params = CP.lateralTuning.torque.as_builder()
self.torque_from_lateral_accel = CI.torque_from_lateral_accel()
self.lateral_accel_from_torque = CI.lateral_accel_from_torque()
self.pid = PIDController([INTERP_SPEEDS, KP_INTERP], KI, rate=1/self.dt)
self.update_limits()
self.steering_angle_deadzone_deg = self.torque_params.steeringAngleDeadzoneDeg
self.lat_accel_request_buffer_len = int(LAT_ACCEL_REQUEST_BUFFER_SECONDS / self.dt)
self.lat_accel_request_buffer = deque([0.] * self.lat_accel_request_buffer_len, maxlen=self.lat_accel_request_buffer_len)
self.lookahead_frames = int(JERK_LOOKAHEAD_SECONDS / self.dt)
self.jerk_filter = FirstOrderFilter(0.0, 1 / (2 * np.pi * LP_FILTER_CUTOFF_HZ), self.dt)
self.previous_measurement = 0.0
self.measurement_rate_filter = FirstOrderFilter(0.0, 1 / (2 * np.pi * (MAX_LAT_JERK_UP - 0.5)), self.dt)
self.low_speed_reset_threshold = max(CP.minSteerSpeed, MIN_LATERAL_CONTROL_SPEED)
self.steer_release_i_decay = 0.8
self.prev_steering_pressed = False
self.debug_counter = 0
self.prev_desired_lateral_accel = 0.0
self.is_bolt = CP.carFingerprint in BOLT_CARS
self.is_bolt_2022_2023 = CP.carFingerprint in BOLT_2022_2023_CARS
self.is_bolt_2018_2021 = CP.carFingerprint in BOLT_2018_2021_CARS
self.is_bolt_2017 = CP.carFingerprint in BOLT_2017_CARS
self.use_bolt_ff_scaling = self.is_bolt_2022_2023 or self.is_bolt_2018_2021 or self.is_bolt_2017
self.use_bolt_ki_multiplier = self.use_bolt_ff_scaling
self.torque_ff_scale_pos = 1.0
self.torque_ff_scale_neg = 1.0
self.torque_deadzone_boost = float(getattr(self.torque_params, "kfDEPRECATED", 0.0))
self.torque_ki_mult = 1.0
if self.is_bolt:
kp_scale = getattr(self.torque_params, "kp", getattr(self.torque_params, "kpDEPRECATED", 1.0))
ki_scale = getattr(self.torque_params, "ki", getattr(self.torque_params, "kiDEPRECATED", 1.0))
kd_scale = getattr(self.torque_params, "kd", getattr(self.torque_params, "kdDEPRECATED", 1.0))
self.torque_ff_scale_pos = float(kp_scale)
self.torque_ff_scale_neg = float(ki_scale)
self.torque_ki_mult = float(kd_scale)
if self.use_bolt_ki_multiplier and self.torque_ki_mult > 0.0 and self.torque_ki_mult != 1.0:
self.pid._k_i = [self.pid._k_i[0], [k * self.torque_ki_mult for k in self.pid._k_i[1]]]
def update_live_torque_params(self, latAccelFactor, latAccelOffset, friction):
self.torque_params.latAccelFactor = latAccelFactor
self.torque_params.latAccelOffset = latAccelOffset
self.torque_params.friction = friction
self.update_limits()
def update_limits(self):
self.pid.set_limits(self.lateral_accel_from_torque(self.steer_max, self.torque_params),
self.lateral_accel_from_torque(-self.steer_max, self.torque_params))
def update(self, active, CS, VM, params, steer_limited_by_safety, desired_curvature, curvature_limited, lat_delay, calibrated_pose, model_data, frogpilot_toggles):
pid_log = log.ControlsState.LateralTorqueState.new_message()
pid_log.version = VERSION
if not active:
output_torque = 0.0
pid_log.active = False
self.pid.reset()
self.previous_measurement = 0.0
self.measurement_rate_filter.x = 0.0
self.lat_accel_request_buffer = deque([0.] * self.lat_accel_request_buffer_len, maxlen=self.lat_accel_request_buffer_len)
self.prev_desired_lateral_accel = 0.0
else:
if self.prev_steering_pressed and not CS.steeringPressed:
self.pid.i *= self.steer_release_i_decay
measured_curvature = -VM.calc_curvature(math.radians(CS.steeringAngleDeg - params.angleOffsetDeg), CS.vEgo, params.roll)
roll_compensation = params.roll * ACCELERATION_DUE_TO_GRAVITY
curvature_deadzone = abs(VM.calc_curvature(math.radians(self.steering_angle_deadzone_deg), CS.vEgo, 0.0))
lateral_accel_deadzone = curvature_deadzone * CS.vEgo ** 2
delay_frames = int(np.clip(lat_delay / self.dt, 1, self.lat_accel_request_buffer_len))
expected_lateral_accel = self.lat_accel_request_buffer[-delay_frames]
future_desired_lateral_accel = desired_curvature * CS.vEgo ** 2
self.lat_accel_request_buffer.append(future_desired_lateral_accel)
raw_lateral_jerk = (future_desired_lateral_accel - expected_lateral_accel) / max(lat_delay, self.dt)
raw_lateral_jerk = np.clip(raw_lateral_jerk, -MAX_LAT_JERK_UP, MAX_LAT_JERK_UP)
desired_lateral_jerk = np.clip(self.jerk_filter.update(raw_lateral_jerk), -MAX_LAT_JERK_UP, MAX_LAT_JERK_UP)
gravity_adjusted_future_lateral_accel = future_desired_lateral_accel - roll_compensation
setpoint = expected_lateral_accel + desired_lateral_jerk * lat_delay
desired_lateral_accel_rate = (setpoint - self.prev_desired_lateral_accel) / self.dt
unwind_detected = (desired_lateral_accel_rate < UNWIND_D_DES_THRESHOLD and
abs(setpoint) < UNWIND_LAT_ACCEL_NEAR_ZERO)
self.prev_desired_lateral_accel = setpoint
measurement = measured_curvature * CS.vEgo ** 2
measurement_rate = self.measurement_rate_filter.update((measurement - self.previous_measurement) / self.dt)
measurement_rate = np.clip(measurement_rate, -MAX_LAT_JERK_UP, MAX_LAT_JERK_UP)
self.previous_measurement = measurement
low_speed_factor = (np.interp(CS.vEgo, LOW_SPEED_X, LOW_SPEED_Y) / max(CS.vEgo, MIN_SPEED)) ** 2
current_kp = np.interp(CS.vEgo, self.pid._k_p[0], self.pid._k_p[1])
error = setpoint - measurement
error_with_lsf = error * (1 + low_speed_factor / max(current_kp, 1e-3))
# do error correction in lateral acceleration space, convert at end to handle non-linear torque responses correctly
pid_log.error = float(error_with_lsf)
ff = gravity_adjusted_future_lateral_accel
# latAccelOffset corrects roll compensation bias from device roll misalignment relative to car roll
ff -= self.torque_params.latAccelOffset
ff_scale = 1.0
if self.use_bolt_ff_scaling:
ff_scale = np.interp(ff, [-FF_SCALE_BLEND_LAT_ACCEL, 0.0, FF_SCALE_BLEND_LAT_ACCEL],
[self.torque_ff_scale_neg, 1.0, self.torque_ff_scale_pos])
ff *= ff_scale
ff += get_friction(error_with_lsf + JERK_GAIN * desired_lateral_jerk, lateral_accel_deadzone, get_friction_threshold(CS.vEgo), self.torque_params)
deadzone_boost_active = False
if self.torque_deadzone_boost > 0.0 and abs(gravity_adjusted_future_lateral_accel) < DEADZONE_BOOST_LAT_ACCEL:
boost_scale = np.interp(abs(gravity_adjusted_future_lateral_accel), [0.0, DEADZONE_BOOST_LAT_ACCEL], [1.0, 0.0])
ff += np.sign(gravity_adjusted_future_lateral_accel) * self.torque_deadzone_boost * boost_scale
deadzone_boost_active = True
if CS.vEgo < self.low_speed_reset_threshold:
self.pid.reset()
freeze_integrator = (steer_limited_by_safety or CS.steeringPressed or
CS.vEgo < self.low_speed_reset_threshold or unwind_detected)
output_lataccel = self.pid.update(pid_log.error, error_rate=-measurement_rate, speed=CS.vEgo, feedforward=ff, freeze_integrator=freeze_integrator)
output_torque = self.torque_from_lateral_accel(output_lataccel, self.torque_params)
pid_log.active = True
pid_log.p = float(self.pid.p)
pid_log.i = float(self.pid.i)
pid_log.d = float(self.pid.d)
pid_log.f = float(self.pid.f)
pid_log.output = float(-output_torque) # TODO: log lat accel?
pid_log.actualLateralAccel = float(measurement)
pid_log.desiredLateralAccel = float(setpoint)
pid_log.desiredLateralJerk = float(desired_lateral_jerk)
pid_log.saturated = bool(self._check_saturation(self.steer_max - abs(output_torque) < 1e-3, CS, steer_limited_by_safety, curvature_limited))
if DEBUG_TORQUE_TUNE and self.is_bolt:
self.debug_counter += 1
if self.debug_counter % 50 == 0:
print(f"bolt_torque ff_scale={ff_scale:.3f} pos={self.torque_ff_scale_pos:.3f} "
f"neg={self.torque_ff_scale_neg:.3f} deadzone_boost_active={deadzone_boost_active}")
self.prev_steering_pressed = CS.steeringPressed
# TODO left is positive in this convention
return -output_torque, 0.0, pid_log