Architecture

Quadrotor Model

The core physics engine is a 6DOF rigid-body quadrotor using Newton-Euler equations with RK4 integration.

State vector (12 dimensions):

• Position: [x, y, z]
• Orientation (Euler): [φ, θ, ψ]
• Linear velocity: [u, v, w]
• Angular velocity: [p, q, r]

Control input (4 dimensions):

• Total thrust T
• Body torques [τ_x, τ_y, τ_z]

Control Stack

The cascaded control architecture follows PX4-inspired design:

Position Controller → Velocity Controller → Attitude Controller → Rate Controller → Motor Mixer

The StateManager implements a finite state machine: DISARMED → ARM → TAKEOFF → HOVER → {TRACKING, OFFBOARD, LOITER} → LAND → DISARMED

Sensors

All sensors simulate real-world noise characteristics:

Sensor	Noise Model	Typical Rate
GPS	Gaussian, 0.5m std	10 Hz
IMU	Gyro bias + white noise	200 Hz
Lidar2D/3D	Range noise + dropout	10-40 Hz
Camera	Projection + distortion	30 Hz

Estimation Pipeline

Filters fuse noisy sensor data into state estimates:

• Complementary Filter: Fast attitude estimation (accel + gyro)
• EKF/UKF: Full 6DOF state estimation with GPS/IMU fusion
• Particle Filter: Non-Gaussian posterior, GPS localisation

Simulation Composition API (Quad-First)

uav_sim.simulations now exposes a composition-first API so users can assemble scenarios without editing internals:

1. Create environment
2. Spawn platform
3. Define mission
4. Run simulation

from pathlib import Path

from uav_sim.simulations import (
    create_environment,
    create_mission,
    create_sim,
    run_sim,
    spawn_quad_platform,
)

env = create_environment(n_buildings=0, world_size=30.0)
platform = spawn_quad_platform()
mission = create_mission(path=my_path, standard=my_standard, fallback_policy="none")
sim = create_sim(name="my_demo", out_dir=Path("."), mission=mission)
result = run_sim(sim=sim, env=env, platform=platform)

Contracts for extensibility

• PayloadPlugin: attach sensors/payload modules that emit standardized packets.
• PathPlannerPlugin, TrackerPlugin, EstimatorPlugin, PerceptionPlugin: pluggable algorithm interfaces.
• QuadPlatform: current reference platform implementation; fixed-wing and VTOL can be added via adapters without changing mission APIs.