Files
gps-denied-onboard/tests/test_eskf.py
T
Yuzviak 78dcf7b4e7 fix: post-audit — runtime bugs, functional gaps, docs, hardening
Phase A — Runtime bugs:
  - SSE: add push_event() method to SSEEventStreamer (was missing, masked by mocks)
  - MAVLink: satellites_visible=10 (was 0, triggers ArduPilot failsafe)
  - MAVLink: horiz_accuracy=sqrt(P[0,0]+P[1,1]) per spec (was sqrt(avg))
  - MAVLink: MEDIUM confidence → fix_type=3 per solution.md (was 2)

Phase B — Functional gaps:
  - handle_user_fix() injects operator GPS into ESKF with noise=500m
  - app.py uses create_vo_backend() factory (was hardcoded SequentialVO)
  - ESKF: Mahalanobis gating on satellite updates (rejects outliers >5σ)
  - ESKF: public accessors (position, quaternion, covariance, last_timestamp)
  - Processor: no more private ESKF field access

Phase C — Documentation:
  - README: correct API endpoints, CLI command, 40+ env vars documented
  - Dockerfile: ENV prefixes match pydantic-settings (DB_, SATELLITE_, MAVLINK_)
  - tech_stack.md marked ARCHIVED (contradicts solution.md)

Phase D — Hardening:
  - JWT auth middleware (AUTH_ENABLED=false default, verify_token on /flights)
  - TLS config env vars (AUTH_SSL_CERTFILE, AUTH_SSL_KEYFILE)
  - SHA-256 tile manifest verification in SatelliteDataManager
  - AuthConfig, ESKFSettings, MAVLinkConfig, SatelliteConfig in config.py

Also: conftest.py shared fixtures, download_tiles.py, convert_to_trt.py scripts,
config wiring into app.py lifespan, config-driven ESKF, calculate_precise_angle fix.

Tests: 196 passed / 8 skipped. Ruff clean.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-02 18:27:35 +03:00

319 lines
11 KiB
Python

"""Tests for ESKF (F17) — Error-State Kalman Filter."""
import numpy as np
import pytest
from gps_denied.core.coordinates import CoordinateTransformer
from gps_denied.core.eskf import ESKF
from gps_denied.schemas import GPSPoint
from gps_denied.schemas.eskf import ConfidenceTier, ESKFConfig, IMUMeasurement
@pytest.fixture
def eskf():
"""ESKF initialized at origin with default config."""
e = ESKF()
e.initialize(np.zeros(3), timestamp=0.0)
return e
@pytest.fixture
def config():
return ESKFConfig()
class TestESKFInitialization:
"""Tests for ESKF-04: Initialization."""
def test_initialization_default(self, config):
"""Test basic initialization with position and timestamp."""
e = ESKF()
e.initialize(np.array([100.0, 200.0, 0.0]), timestamp=0.0)
assert e.initialized
state = e.get_state()
assert np.allclose(state.position, [100.0, 200.0, 0.0])
assert np.allclose(state.velocity, [0.0, 0.0, 0.0])
assert np.allclose(state.quaternion, [1.0, 0.0, 0.0, 0.0])
assert state.covariance.shape == (15, 15)
assert state.covariance[0, 0] == config.init_pos_var
assert state.covariance[3, 3] == config.init_vel_var
assert state.covariance[6, 6] == config.init_att_var
def test_initialization_from_gps(self):
"""Test initialization from GPS position."""
ct = CoordinateTransformer()
ct.set_enu_origin("f1", GPSPoint(lat=48.0, lon=37.0))
e = ESKF()
e.initialize_from_gps(
GPSPoint(lat=48.001, lon=37.001),
altitude=600.0,
timestamp=0.0,
coord_transformer=ct,
flight_id="f1",
)
assert e.initialized
state = e.get_state()
assert not np.allclose(state.position, [0.0, 0.0, 0.0])
assert np.linalg.norm(state.position[:2]) > 50 # ~111m per 0.001 deg
class TestESKFPrediction:
"""Tests for ESKF-01: IMU prediction."""
def test_predict_covariance_grows(self, eskf):
"""Test that covariance grows during IMU-only prediction."""
P_before = eskf.get_state().covariance.copy()
trace_before = np.trace(P_before[0:3, 0:3])
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01,
)
eskf.predict(imu)
P_after = eskf.get_state().covariance
trace_after = np.trace(P_after[0:3, 0:3])
assert trace_after > trace_before
def test_predict_gravity_compensation(self, eskf):
"""Test that gravity is properly compensated (stationary accel has no velocity growth)."""
# 100 steps at 0.01s = 1 second with gravity-aligned acceleration
for i in range(100):
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
eskf.predict(imu)
state = eskf.get_state()
# Velocity should remain near zero (gravity compensated)
assert np.linalg.norm(state.velocity) < 1.0
def test_predict_with_acceleration(self, eskf):
"""Test velocity integration from applied acceleration."""
# 100 steps at 0.01s = 1 second with 1 m/s^2 forward + gravity
for i in range(100):
imu = IMUMeasurement(
accel=np.array([1.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
eskf.predict(imu)
state = eskf.get_state()
# Velocity[0] should be approximately 1.0 m/s
assert abs(state.velocity[0] - 1.0) < 0.5
def test_predict_position_propagation(self):
"""Test position propagation from velocity."""
e = ESKF()
e.initialize(np.zeros(3), timestamp=0.0, velocity=np.array([10.0, 0.0, 0.0]))
for i in range(100):
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
e.predict(imu)
state = e.get_state()
# Position[0] should be approximately 10.0 meters after 1 second at 10 m/s
assert abs(state.position[0] - 10.0) < 2.0
class TestESKFVOUpdate:
"""Tests for ESKF-02: VO measurement update."""
def test_vo_update_reduces_uncertainty(self, eskf):
"""Test that VO update reduces position covariance."""
# Grow covariance first
for _ in range(5):
imu = IMUMeasurement(
accel=np.zeros(3),
gyro=np.zeros(3),
timestamp=eskf._last_timestamp + 0.01,
)
eskf.predict(imu)
trace_before = np.trace(eskf.get_state().covariance[0:3, 0:3])
eskf.update_vo(np.zeros(3), dt_vo=0.1)
trace_after = np.trace(eskf.get_state().covariance[0:3, 0:3])
assert trace_after < trace_before
def test_vo_update_corrects_position(self, eskf):
"""Test that VO update shifts position toward measurement."""
# Predict to drift
for i in range(10):
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
eskf.predict(imu)
pos_before = eskf.get_state().position.copy()
# VO measurement with 1m north offset
eskf.update_vo(np.array([0.0, 1.0, 0.0]), dt_vo=0.1)
pos_after = eskf.get_state().position
assert not np.allclose(pos_after, pos_before)
def test_vo_update_returns_innovation(self, eskf):
"""Test that VO update returns innovation vector."""
innovation = eskf.update_vo(np.array([1.0, 0.0, 0.0]), dt_vo=0.1)
assert innovation.shape == (3,)
assert np.linalg.norm(innovation) > 0.5
class TestESKFSatelliteUpdate:
"""Tests for ESKF-03: Satellite measurement update."""
def test_satellite_update_corrects_position(self, eskf):
"""Test that satellite update moves position toward measurement."""
# Predict to drift
for i in range(10):
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
eskf.predict(imu)
pos_before = eskf.get_state().position.copy()
# Use target within Mahalanobis gate (small shift relative to P)
target_pos = np.array([5.0, 5.0, 0.0])
eskf.update_satellite(target_pos, noise_meters=10.0)
pos_after = eskf.get_state().position
# Position should move toward target
dist_before = np.linalg.norm(pos_before - target_pos)
dist_after = np.linalg.norm(pos_after - target_pos)
assert dist_after < dist_before
def test_satellite_update_tightens_covariance(self, eskf):
"""Test that satellite update reduces position covariance."""
trace_before = np.trace(eskf.get_state().covariance[0:3, 0:3])
eskf.update_satellite(np.zeros(3), noise_meters=5.0)
trace_after = np.trace(eskf.get_state().covariance[0:3, 0:3])
assert trace_after < trace_before
def test_satellite_update_covariance_bounded_by_noise(self, eskf):
"""Test that final covariance is bounded by satellite noise level."""
eskf.update_satellite(np.zeros(3), noise_meters=5.0)
pos_cov_trace = np.trace(eskf.get_state().covariance[0:3, 0:3])
# Covariance should be bounded by roughly 3*noise^2 (3 dimensions)
assert pos_cov_trace < 3 * (5.0 ** 2)
class TestESKFConfidenceTiers:
"""Tests for ESKF-05: Confidence tier computation."""
def test_confidence_high(self, eskf, config):
"""Test HIGH confidence with recent satellite and low covariance."""
eskf.update_satellite(np.zeros(3), noise_meters=5.0)
# Advance time slightly (still within satellite_max_age)
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=1.0,
)
eskf.predict(imu)
assert eskf.get_confidence() == ConfidenceTier.HIGH
def test_confidence_medium(self):
"""Test MEDIUM confidence with recent VO but no satellite."""
e = ESKF()
e.initialize(np.zeros(3), timestamp=0.0)
e.update_vo(np.zeros(3), dt_vo=0.1)
assert e.get_confidence() == ConfidenceTier.MEDIUM
def test_confidence_low(self, eskf):
"""Test LOW confidence with no recent measurements."""
# Many predictions without any update
for i in range(100):
imu = IMUMeasurement(
accel=np.zeros(3),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
eskf.predict(imu)
assert eskf.get_confidence() == ConfidenceTier.LOW
def test_confidence_failed(self, eskf):
"""Test FAILED confidence with 3+ consecutive failures."""
assert eskf.get_confidence(consecutive_failures=3) == ConfidenceTier.FAILED
assert eskf.get_confidence(consecutive_failures=0) != ConfidenceTier.FAILED
class TestESKFStateAccess:
"""Tests for ESKF state accessor."""
def test_get_state_returns_eskf_state(self, eskf):
"""Test that get_state returns properly formed ESKFState."""
state = eskf.get_state()
assert state.position.shape == (3,)
assert state.velocity.shape == (3,)
assert state.quaternion.shape == (4,)
assert state.accel_bias.shape == (3,)
assert state.gyro_bias.shape == (3,)
assert state.covariance.shape == (15, 15)
assert isinstance(state.confidence, ConfidenceTier)
class TestESKFFusion:
"""Integration tests for full ESKF fusion sequence."""
def test_full_fusion_sequence(self):
"""Test complete IMU → VO → satellite fusion."""
e = ESKF()
e.initialize(np.zeros(3), timestamp=0.0)
# IMU prediction phase
for i in range(10):
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 1),
)
e.predict(imu)
P_before_vo = np.trace(e.get_state().covariance[0:3, 0:3])
# VO update
e.update_vo(np.array([0.1, 0.0, 0.0]), dt_vo=0.1)
P_after_vo = np.trace(e.get_state().covariance[0:3, 0:3])
assert P_after_vo < P_before_vo
# More IMU prediction
for i in range(10):
imu = IMUMeasurement(
accel=np.array([0.0, 0.0, 9.81]),
gyro=np.zeros(3),
timestamp=0.01 * (i + 11),
)
e.predict(imu)
# Satellite update
e.update_satellite(np.array([1.0, 0.0, 0.0]), noise_meters=5.0)
state = e.get_state()
# Position should be close to satellite measurement
assert np.linalg.norm(state.position[:2] - [1.0, 0.0]) < 10.0
assert state.confidence == ConfidenceTier.HIGH