#!/usr/bin/env python3 import math import numpy as np import time from openpilot.common.numpy_fast import clip, interp import cereal.messaging as messaging from openpilot.common.conversions import Conversions as CV from openpilot.common.filter_simple import FirstOrderFilter from openpilot.common.realtime import DT_MDL from openpilot.selfdrive.modeld.constants import ModelConstants from openpilot.selfdrive.car.interfaces import ACCEL_MIN, ACCEL_MAX from openpilot.selfdrive.controls.lib.longcontrol import LongCtrlState from openpilot.selfdrive.controls.lib.longitudinal_mpc_lib.long_mpc import LongitudinalMpc from openpilot.selfdrive.controls.lib.longitudinal_mpc_lib.long_mpc import T_IDXS as T_IDXS_MPC from openpilot.selfdrive.controls.lib.drive_helpers import V_CRUISE_UNSET, CONTROL_N, get_speed_error, get_accel_from_plan_tomb_raider from openpilot.common.swaglog import cloudlog LON_MPC_STEP = 0.2 # first step is 0.2s A_CRUISE_MIN = -1.0 A_CRUISE_MAX_BP = [0.0, 5., 10., 15., 20., 25., 40.] A_CRUISE_MAX_VALS = [1.125, 1.125, 1.125, 1.125, 1.25, 1.25, 1.5] CONTROL_N_T_IDX = ModelConstants.T_IDXS[:CONTROL_N] ALLOW_THROTTLE_THRESHOLD = 0.5 MIN_ALLOW_THROTTLE_SPEED = 2.5 # Uncertainty-based filter disable thresholds UNCERT_SLOPE_TRIG = 0.12 # per second UNCERT_MAG_TRIG = 0.50 # Lookup table for turns _A_TOTAL_MAX_V = [1.7, 3.2] _A_TOTAL_MAX_BP = [20., 40.] def get_max_accel(v_ego): return interp(v_ego, A_CRUISE_MAX_BP, A_CRUISE_MAX_VALS) def get_coast_accel(pitch): return np.sin(pitch) * -5.65 - 0.3 # fitted from data using xx/projects/allow_throttle/compute_coast_accel.py def limit_accel_in_turns(v_ego, angle_steers, a_target, CP): """ This function returns a limited long acceleration allowed, depending on the existing lateral acceleration this should avoid accelerating when losing the target in turns """ # FIXME: This function to calculate lateral accel is incorrect and should use the VehicleModel # The lookup table for turns should also be updated if we do this a_total_max = interp(v_ego, _A_TOTAL_MAX_BP, _A_TOTAL_MAX_V) a_y = v_ego ** 2 * angle_steers * CV.DEG_TO_RAD / (CP.steerRatio * CP.wheelbase) a_x_allowed = math.sqrt(max(a_total_max ** 2 - a_y ** 2, 0.)) return [a_target[0], min(a_target[1], a_x_allowed)] def get_accel_from_plan_classic(CP, speeds, accels, vEgoStopping): if len(speeds) == CONTROL_N: v_target_now = interp(DT_MDL, CONTROL_N_T_IDX, speeds) a_target_now = interp(DT_MDL, CONTROL_N_T_IDX, accels) v_target = interp(CP.longitudinalActuatorDelay + DT_MDL, CONTROL_N_T_IDX, speeds) if v_target != v_target_now: a_target = 2 * (v_target - v_target_now) / CP.longitudinalActuatorDelay - a_target_now else: a_target = a_target_now v_target_1sec = interp(CP.longitudinalActuatorDelay + DT_MDL + 1.0, CONTROL_N_T_IDX, speeds) else: v_target = 0.0 v_target_1sec = 0.0 a_target = 0.0 should_stop = (v_target < vEgoStopping and v_target_1sec < vEgoStopping) return a_target, should_stop def get_accel_from_plan(speeds, accels, action_t=DT_MDL, vEgoStopping=0.05): if len(speeds) == CONTROL_N: v_now = speeds[0] a_now = accels[0] v_target = interp(action_t, CONTROL_N_T_IDX, speeds) a_target = 2 * (v_target - v_now) / (action_t) - a_now v_target_1sec = interp(action_t + 1.0, CONTROL_N_T_IDX, speeds) else: v_target = 0.0 v_target_1sec = 0.0 a_target = 0.0 should_stop = (v_target < vEgoStopping and v_target_1sec < vEgoStopping) return a_target, should_stop class LongitudinalPlanner: def __init__(self, CP, init_v=0.0, init_a=0.0, dt=DT_MDL): self.CP = CP self.mpc = LongitudinalMpc(dt=dt) self.fcw = False self.dt = dt self.allow_throttle = True self.mode = 'acc' self.generation = None self.a_desired = init_a self.v_desired_filter = FirstOrderFilter(init_v, 2.0, self.dt) self.v_model_error = 0.0 self.v_desired_trajectory = np.zeros(CONTROL_N) self.a_desired_trajectory = np.zeros(CONTROL_N) self.j_desired_trajectory = np.zeros(CONTROL_N) self.solverExecutionTime = 0.0 # logging cadence & state self.last_uncert_log_t = 0.0 self.prev_uncert_over = False # ---- Rubberband mitigation state ---- # Two uncertainty tracks (slow/fast) for asymmetric gating self.uncert_slow = FirstOrderFilter(0.0, 1.6, self.dt) # ~lam=0.6 self.uncert_fast = FirstOrderFilter(0.0, 0.9, self.dt) # faster cool-down for accel decisions # Lead stability tracking self.prev_lead_dist = None self.last_big_brake_t = 0.0 self.stable_lead = False # Temporary accel nudge window self.accel_nudge_until = 0.0 # Hysteresis gate + dwell for accel re-engage and smoothed lead distance self.accel_gate = False self._t_arm = 0.0 self._t_disarm = 0.0 self.lead_dist_f = None # Uncertainty slope tracking self._uncert_last = 0.0 self._uncert_last_t = None @property def mlsim(self): return self.generation in ("v8", "v10", "v11") def get_mpc_mode(self) -> str: """ Determine the desired MPC mode: if not ML-SIM, MPC should follow self.mode; otherwise leave MPC.mode unchanged. """ # For non-ML-SIM generations, MPC mode tracks self.mode if not self.mlsim: return self.mode # For ML-SIM (v8), preserve the existing MPC mode return getattr(self.mpc, 'mode', 'acc') @staticmethod def parse_model(model_msg, model_error, v_ego, taco_tune): if (len(model_msg.position.x) == ModelConstants.IDX_N and len(model_msg.velocity.x) == ModelConstants.IDX_N and len(model_msg.acceleration.x) == ModelConstants.IDX_N): x = np.interp(T_IDXS_MPC, ModelConstants.T_IDXS, model_msg.position.x) - model_error * T_IDXS_MPC v = np.interp(T_IDXS_MPC, ModelConstants.T_IDXS, model_msg.velocity.x) - model_error a = np.interp(T_IDXS_MPC, ModelConstants.T_IDXS, model_msg.acceleration.x) j = np.zeros(len(T_IDXS_MPC)) else: x = np.zeros(len(T_IDXS_MPC)) v = np.zeros(len(T_IDXS_MPC)) a = np.zeros(len(T_IDXS_MPC)) j = np.zeros(len(T_IDXS_MPC)) if taco_tune: max_lat_accel = interp(v_ego, [5, 10, 20], [1.5, 2.0, 3.0]) curvatures = np.interp(T_IDXS_MPC, ModelConstants.T_IDXS, model_msg.orientationRate.z) / np.clip(v, 0.3, 100.0) max_v = np.sqrt(max_lat_accel / (np.abs(curvatures) + 1e-3)) - 2.0 v = np.minimum(max_v, v) if len(model_msg.meta.disengagePredictions.gasPressProbs) > 1: throttle_prob = model_msg.meta.disengagePredictions.gasPressProbs[1] else: throttle_prob = 1.0 return x, v, a, j, throttle_prob def update(self, tinygrad_model, sm, frogpilot_toggles): self.generation = frogpilot_toggles.model_version if tinygrad_model: self.mpc.mode = 'acc' self.mode = 'blended' if sm['controlsState'].experimentalMode else 'acc' else: self.mpc.mode = 'blended' if sm['controlsState'].experimentalMode else 'acc' if not self.mlsim: self.mpc.mode = self.mode if len(sm['carControl'].orientationNED) == 3: accel_coast = get_coast_accel(sm['carControl'].orientationNED[1]) else: accel_coast = ACCEL_MAX v_ego = max(sm['carState'].vEgo, sm['carState'].vEgoCluster) v_cruise = sm['frogpilotPlan'].vCruise v_cruise_initialized = sm['controlsState'].vCruise != V_CRUISE_UNSET long_control_off = sm['controlsState'].longControlState == LongCtrlState.off force_slow_decel = sm['controlsState'].forceDecel # Reset current state when not engaged, or user is controlling the speed reset_state = long_control_off if self.CP.openpilotLongitudinalControl else not sm['controlsState'].enabled # PCM cruise speed may be updated a few cycles later, check if initialized reset_state = reset_state or not v_cruise_initialized # No change cost when user is controlling the speed, or when standstill prev_accel_constraint = not (reset_state or sm['carState'].standstill) if self.mpc.mode == 'acc': accel_limits = [sm['frogpilotPlan'].minAcceleration, sm['frogpilotPlan'].maxAcceleration] steer_angle_without_offset = sm['carState'].steeringAngleDeg - sm['liveParameters'].angleOffsetDeg accel_limits_turns = limit_accel_in_turns(v_ego, steer_angle_without_offset, accel_limits, self.CP) else: accel_limits = [ACCEL_MIN, ACCEL_MAX] accel_limits_turns = [ACCEL_MIN, ACCEL_MAX] if reset_state: self.v_desired_filter.x = v_ego # Clip aEgo to cruise limits to prevent large accelerations when becoming active self.a_desired = clip(sm['carState'].aEgo, accel_limits[0], accel_limits[1]) # Prevent divergence, smooth in current v_ego self.v_desired_filter.x = max(0.0, self.v_desired_filter.update(v_ego)) # Compute model v_ego error self.v_model_error = get_speed_error(sm['modelV2'], v_ego) x, v, a, j, throttle_prob = self.parse_model(sm['modelV2'], self.v_model_error, v_ego, frogpilot_toggles.taco_tune) # Don't clip at low speeds since throttle_prob doesn't account for creep self.allow_throttle = throttle_prob > ALLOW_THROTTLE_THRESHOLD or v_ego <= MIN_ALLOW_THROTTLE_SPEED self.allow_throttle &= not sm['frogpilotPlan'].disableThrottle if not self.allow_throttle: clipped_accel_coast = max(accel_coast, accel_limits_turns[0]) clipped_accel_coast_interp = interp(v_ego, [MIN_ALLOW_THROTTLE_SPEED, MIN_ALLOW_THROTTLE_SPEED*2], [accel_limits_turns[1], clipped_accel_coast]) accel_limits_turns[1] = min(accel_limits_turns[1], clipped_accel_coast_interp) if force_slow_decel: v_cruise = 0.0 # clip limits, cannot init MPC outside of bounds accel_limits_turns[0] = min(accel_limits_turns[0], self.a_desired + 0.05) accel_limits_turns[1] = max(accel_limits_turns[1], self.a_desired - 0.05) self.lead_one = sm['radarState'].leadOne self.lead_two = sm['radarState'].leadTwo lead_dist = self.lead_one.dRel if self.lead_one.status else 50.0 # Smooth lead distance (EMA) to avoid chatter in thresholds alpha = max(0.02, min(0.15, 0.05 + 0.002 * v_ego)) if self.lead_dist_f is None: self.lead_dist_f = float(lead_dist) else: self.lead_dist_f += alpha * (float(lead_dist) - self.lead_dist_f) # Lead stability estimation and recent-brake timer now_t = time.monotonic() # relative speed (ego - lead) positive when closing v_rel = (v_ego - self.lead_one.vLead) if self.lead_one.status else 0.0 if self.prev_lead_dist is None: d_rel_dot = 0.0 else: d_rel_dot = (lead_dist - self.prev_lead_dist) / max(self.dt, 1e-3) self.prev_lead_dist = lead_dist # Remember time of last non-trivial model brake risk if 'raw_brake_max' in locals() and raw_brake_max is not None and raw_brake_max > 0.02: self.last_big_brake_t = now_t # Stable lead heuristic (short window, cheap to compute) recently_braked = (now_t - self.last_big_brake_t) < 0.7 self.stable_lead = ( self.lead_one.status and abs(v_rel) < 0.5 and abs(d_rel_dot) < 0.5 and not recently_braked ) # Calculate scene uncertainty from model desire prediction entropy and disengage predictions uncertainty = 0.0 if hasattr(sm['modelV2'], 'meta'): # Desire prediction entropy (maneuver uncertainty), normalized to [0, 1] desire_entropy = 0.0 if hasattr(sm['modelV2'].meta, 'desirePrediction'): desire_probs = sm['modelV2'].meta.desirePrediction if len(desire_probs) > 1: probs = np.asarray(desire_probs, dtype=float) total = float(np.sum(probs)) if total > 1e-6: p = probs / total entropy = -np.sum(p * np.log(p + 1e-10)) max_entropy = np.log(len(p)) desire_entropy = float(entropy / max(max_entropy, 1e-6)) # normalized entropy in [0,1] else: desire_entropy = 0.0 # guard against all-zero vector # Disengage prediction risk (intervention likelihood) disengage_risk = 0.0 raw_brake_max = -1.0 lam = -1.0 if hasattr(sm['modelV2'].meta, 'disengagePredictions'): # Use brake press probabilities as primary risk indicator brake_probs = sm['modelV2'].meta.disengagePredictions.brakePressProbs if len(brake_probs) > 0: # Exponentially decayed max over the full horizon probs = np.asarray(brake_probs, dtype=float) # Clip tiny brake blips so they don't inflate uncertainty if float(np.max(probs)) < 0.015: probs = probs * 0.5 raw_brake_max = float(np.max(probs)) # Time vector assuming model horizon step = DT_MDL t = np.arange(len(probs), dtype=float) * DT_MDL lam = 0.6 # decay rate per second (tunable: 0.5–0.9 typical) weights = np.exp(-lam * t) disengage_risk = float(np.max(probs * weights)) # Combined uncertainty metric (range roughly 0..2), with dual-track filtering raw_uncertainty = desire_entropy + disengage_risk # Update filters self.uncert_slow.update(raw_uncertainty) self.uncert_fast.update(raw_uncertainty) # Use a more permissive track for accel decisions uncertainty = self.uncert_slow.x uncertainty_accel = min(self.uncert_slow.x, self.uncert_fast.x) # --- Slope-based panic bypass --- if self._uncert_last_t is None: uncert_slope = 0.0 else: dt_u = max(1e-3, now_t - self._uncert_last_t) uncert_slope = (uncertainty - self._uncert_last) / dt_u self._uncert_last = uncertainty self._uncert_last_t = now_t closing_fast = (self.lead_one.status and (v_ego - self.lead_one.vLead) > 0.5) # Trigger if either slope is high or magnitude is high; require a valid lead and closing panic_bypass = closing_fast and (uncert_slope > UNCERT_SLOPE_TRIG or uncertainty >= UNCERT_MAG_TRIG) if panic_bypass: try: cloudlog.error(f"LON_SLOPE; slope={uncert_slope:.3f}/s; uncertainty={uncertainty:.3f}; v_ego={v_ego:.2f}; v_rel={(v_ego - self.lead_one.vLead) if self.lead_one.status else 0.0:.2f}; lead_dist={self.lead_dist_f if self.lead_dist_f is not None else -1:.2f}; trigger=True") except Exception: pass # now_t defined earlier over = uncertainty > 1.0 # Log on threshold edge or at ~1 Hz if over != self.prev_uncert_over or (now_t - self.last_uncert_log_t) > 1.0: try: cloudlog.error( f"LON_UNCERT; v_ego={v_ego:.2f} mps; desireEntropy={desire_entropy:.3f}; " f"brakeRawMax={(raw_brake_max if 'raw_brake_max' in locals() else -1.0):.3f}; " f"brakeDecayed={(disengage_risk if 'disengage_risk' in locals() else -1.0):.3f}; " f"lam={(lam if 'lam' in locals() else -1.0):.2f}; uncertainty={uncertainty:.3f}; over={over}" ) except Exception as e: cloudlog.warning(f"LON_UNCERT log error: {e}") self.prev_uncert_over = over self.last_uncert_log_t = now_t # Asymmetric accel release with hysteresis + dwell to prevent on/off pulsing rise_dwell_s, fall_dwell_s = 0.6, 0.4 good = ( (self.a_desired > 0.0) and self.stable_lead and (uncertainty <= 0.425) and (desire_entropy < 0.41) ) # dwell timers for robust gating if good and not self.accel_gate: if now_t - self._t_arm >= rise_dwell_s: self.accel_gate = True else: self._t_arm = now_t if (not good) and self.accel_gate: if now_t - self._t_disarm >= fall_dwell_s: self.accel_gate = False else: self._t_disarm = now_t if self.accel_gate: # Ensure some positive headroom for MPC to exit coasting accel_limits_turns[1] = max(accel_limits_turns[1], 0.2) # Short self-canceling nudge to unstick (applied post-MPC) if now_t > self.accel_nudge_until: self.accel_nudge_until = now_t + 0.45 self.mpc.set_weights(sm['frogpilotPlan'].accelerationJerk, sm['frogpilotPlan'].dangerJerk, sm['frogpilotPlan'].speedJerk, prev_accel_constraint, personality=sm['controlsState'].personality, v_ego=v_ego, lead_dist=self.lead_dist_f if self.lead_dist_f is not None else lead_dist, uncertainty=uncertainty, accel_reengage=self.accel_gate, panic_bypass=panic_bypass) self.mpc.set_accel_limits(accel_limits_turns[0], accel_limits_turns[1]) self.mpc.set_cur_state(self.v_desired_filter.x, self.a_desired) # After deciding the MPC mode via get_mpc_mode(), ensure MPC uses that mode when not mlsim dec_mpc_mode = self.get_mpc_mode() if not self.mlsim: self.mpc.mode = dec_mpc_mode self.mpc.update(self.lead_one, self.lead_two, v_cruise, x, v, a, j, sm['frogpilotPlan'].tFollow, sm['frogpilotPlan'].trackingLead, personality=sm['controlsState'].personality) self.a_desired_trajectory_full = np.interp(CONTROL_N_T_IDX, T_IDXS_MPC, self.mpc.a_solution) self.v_desired_trajectory = np.interp(CONTROL_N_T_IDX, T_IDXS_MPC, self.mpc.v_solution) self.a_desired_trajectory = np.interp(CONTROL_N_T_IDX, T_IDXS_MPC, self.mpc.a_solution) self.j_desired_trajectory = np.interp(CONTROL_N_T_IDX, T_IDXS_MPC[:-1], self.mpc.j_solution) # TODO counter is only needed because radar is glitchy, remove once radar is gone self.fcw = self.mpc.crash_cnt > 2 and not sm['carState'].standstill if self.fcw: cloudlog.info("FCW triggered") # Safety checks for rubber-banding mitigation max_jerk = np.max(np.abs(self.mpc.j_solution)) max_accel_change = np.max(np.abs(np.diff(self.mpc.a_solution))) if max_jerk > 5.0: # m/s^3 cloudlog.warning(f"High jerk detected: {max_jerk:.2f} m/s^3") if max_accel_change > 2.0: # m/s^2 cloudlog.warning(f"High acceleration change: {max_accel_change:.2f} m/s^2") # Interpolate 0.05 seconds and save as starting point for next iteration a_prev = self.a_desired self.a_desired = float(interp(self.dt, CONTROL_N_T_IDX, self.a_desired_trajectory)) self.v_desired_filter.x = self.v_desired_filter.x + self.dt * (self.a_desired + a_prev) / 2.0 # Anticipatory pre-brake to avoid "coming in hot" when closing on a lead if self.lead_one.status: rel_v = max(0.0, v_ego - self.lead_one.vLead) # dynamic time headway adds a small buffer when uncertainty is elevated base_th = 1.6 th = base_th + 0.6 * max(0.0, uncertainty - 0.42) desired_gap = th * v_ego if (self.lead_dist_f is not None and self.lead_dist_f < desired_gap and rel_v > 0.5): k_rel, k_unc = 0.04, 0.20 pre_brake = k_rel * rel_v + k_unc * max(0.0, uncertainty - 0.42) pre_brake = min(pre_brake, 0.06) self.a_desired = float(self.a_desired - pre_brake) # Apply tiny feed-forward nudge when released and safe if now_t < self.accel_nudge_until and self.a_desired > -0.1: self.a_desired = float(min(self.a_desired + 0.12, get_max_accel(v_ego))) # Small deadzone around zero accel to kill micro-dithers if -0.05 < self.a_desired < 0.05: self.a_desired = 0.0 def publish(self, classic_model, tinygrad_model, sm, pm, frogpilot_toggles): plan_send = messaging.new_message('longitudinalPlan') plan_send.valid = sm.all_checks(service_list=['carState', 'controlsState']) longitudinalPlan = plan_send.longitudinalPlan longitudinalPlan.modelMonoTime = sm.logMonoTime['modelV2'] longitudinalPlan.processingDelay = (plan_send.logMonoTime / 1e9) - sm.logMonoTime['modelV2'] longitudinalPlan.solverExecutionTime = self.mpc.solve_time longitudinalPlan.speeds = self.v_desired_trajectory.tolist() longitudinalPlan.accels = self.a_desired_trajectory.tolist() longitudinalPlan.jerks = self.j_desired_trajectory.tolist() longitudinalPlan.hasLead = self.lead_one.status longitudinalPlan.longitudinalPlanSource = self.mpc.source longitudinalPlan.fcw = self.fcw if classic_model: a_target, should_stop = get_accel_from_plan_classic(self.CP, longitudinalPlan.speeds, longitudinalPlan.accels, vEgoStopping=frogpilot_toggles.vEgoStopping) elif tinygrad_model: action_t = self.CP.longitudinalActuatorDelay + DT_MDL output_a_target_mpc, output_should_stop_mpc = get_accel_from_plan_tomb_raider(self.v_desired_trajectory, self.a_desired_trajectory, CONTROL_N_T_IDX, action_t=action_t, vEgoStopping=frogpilot_toggles.vEgoStopping) output_a_target_e2e = sm['modelV2'].action.desiredAcceleration output_should_stop_e2e = sm['modelV2'].action.shouldStop # v9 uses a different longitudinal interface; keep MPC-only behavior even in blended mode if self.mode == 'acc' or self.generation == 'v9': a_target = output_a_target_mpc should_stop = output_should_stop_mpc else: a_target = min(output_a_target_mpc, output_a_target_e2e) should_stop = output_should_stop_e2e or output_should_stop_mpc else: action_t = self.CP.longitudinalActuatorDelay + DT_MDL a_target, should_stop = get_accel_from_plan(longitudinalPlan.speeds, longitudinalPlan.accels, action_t=action_t, vEgoStopping=frogpilot_toggles.vEgoStopping) longitudinalPlan.aTarget = float(a_target) longitudinalPlan.shouldStop = bool(should_stop) or sm['frogpilotPlan'].forcingStopLength < 1 longitudinalPlan.allowBrake = True longitudinalPlan.allowThrottle = self.allow_throttle pm.send('longitudinalPlan', plan_send)