Sensor Mounting

Sensor mounting positions and alignment for the AV2.

Overview 

Accurate sensor mounting and calibration is critical for:

Correct coordinate transformations
Accurate obstacle localization
Proper sensor fusion
Safe autonomous operation

All sensor positions are defined relative to the vehicle frame origin (rear axle center at ground level).

Sensor Positions Summary 

                TOP VIEW
┌─────────────────────────────────────┐
│             ROOF                    │
│                                     │
│        ┌─────────────┐              │
│        │   LIDAR     │ ◄── Velodyne │
│        │  (0.3, 0, 1.8)             │
│        └─────────────┘              │
│                                     │
│        ┌─────────────┐              │
│        │   GPS/IMU   │ ◄── Xsens    │
│        │  (0.0, 0, 1.7)             │
│        └─────────────┘              │
│                                     │
├─────────────────────────────────────┤
│                                     │
│  ┌───────────────────────────────┐  │
│  │        WINDSHIELD             │  │
│  │                               │  │
│  │      ┌───────┐                │  │
│  │      │CAMERA │ ◄── RealSense  │  │
│  │      │(1.5, 0, 1.2)           │  │
│  │      └───────┘                │  │
│  │                               │  │
│  └───────────────────────────────┘  │
│                                     │
│                                     │
│  ┌───────────────────────────────┐  │
│  │           HOOD                │  │
│  └───────────────────────────────┘  │
│                                     │
│  X ◄── Vehicle Frame Origin         │
│       (Rear Axle Center)            │
│                                     │
└─────────────────────────────────────┘
               FRONT

Position Format: (X, Y, Z) in meters X = forward, Y = left, Z = up

Velodyne VLP-16 (LIDAR)

Mounting Position 

Axis	Value	Notes
X (forward)	0.3 m	Slightly forward
Y (left)	0.0 m	Centered
Z (up)	1.8 m	Roof mounted

Orientation 

Angle	Value	Notes
Roll	0°	Level
Pitch	0°	Horizontal
Yaw	0°	X-axis forward

Mounting Requirements 

Level: Use bubble level to ensure horizontal
Clear FOV: No obstructions in 360° horizontal plane
Vibration Isolation: Use rubber mounts to reduce vibration
Cable Routing: Protect Ethernet and power cables

Field of View Clearance 

The LIDAR has ±15° vertical FOV. Ensure no obstructions:

At 1.8m height with ±15° vertical FOV:

Upper limit: tan(15°) × distance
Lower limit: tan(-15°) × distance

At 10m range:
- Upper ray: 1.8 + 2.68 = 4.48m above ground
- Lower ray: 1.8 - 2.68 = -0.88m (below ground = hits ground at ~6.7m)

Xsens MTi-630 (GPS/IMU)

Mounting Position 

Axis	Value	Notes
X (forward)	0.0 m	At rear axle
Y (left)	0.0 m	Centered
Z (up)	1.7 m	Roof mounted

Note

Mounting at the rear axle center simplifies coordinate transformations since the vehicle frame origin is defined here.

Orientation 

Angle	Value	Notes
Roll	0°	Level
Pitch	0°	Horizontal
Yaw	0°	X-axis forward

Mounting Requirements 

Rigid Mount: No flexing or vibration
Alignment: X-axis must point exactly forward
Level: Critical for accurate orientation
GPS Antenna: Clear sky view required

GPS Antenna Placement 

If using external GPS antenna:

Mount on highest point of vehicle
Metal ground plane (roof) improves reception
Keep away from other antennas
Minimum 10cm from edges

Intel RealSense D435 (Camera)

Mounting Position 

Axis	Value	Notes
X (forward)	1.5 m	Front of vehicle
Y (left)	0.0 m	Centered
Z (up)	1.2 m	Windshield level

Orientation 

Angle	Value	Notes
Roll	0°	Level
Pitch	-5° to -15°	Slight downward
Yaw	0°	Forward facing

Note

A slight downward pitch (5-15°) helps capture the road surface while still seeing distant obstacles.

Mounting Requirements 

Stable Mount: Minimize vibration
Clean View: No reflections from windshield
Avoid Sun: Direct sunlight can blind IR sensors
USB Cable: High-quality USB 3.0, < 3m length

Calibration 

Extrinsic Calibration 

Sensor-to-vehicle transformations are defined in the configuration:

# Example: LIDAR to vehicle transform
lidar_to_vehicle = {
    'translation': [0.3, 0.0, 1.8],  # [x, y, z] meters
    'rotation': [0, 0, 0]             # [roll, pitch, yaw] degrees
}

# Example: Camera to vehicle transform
camera_to_vehicle = {
    'translation': [1.5, 0.0, 1.2],
    'rotation': [0, -10, 0]  # 10° down pitch
}

Calibration Procedure 

Measure Physical Positions: - Use tape measure from rear axle center - Record X, Y, Z for each sensor
Verify Alignment: - Use laser level for roll/pitch - Use reference line for yaw
Fine-Tune with Data: - Collect sample data - Verify point cloud alignment - Adjust offsets as needed

Transformation Matrices 

The full transformation from sensor frame to vehicle frame:

import numpy as np

def sensor_to_vehicle_transform(translation, rotation_deg):
    """Create 4x4 homogeneous transformation matrix."""
    roll, pitch, yaw = np.radians(rotation_deg)

    # Rotation matrices
    Rx = rotation_matrix_x(roll)
    Ry = rotation_matrix_y(pitch)
    Rz = rotation_matrix_z(yaw)

    R = Rz @ Ry @ Rx

    T = np.eye(4)
    T[:3, :3] = R
    T[:3, 3] = translation

    return T

See Coordinate Systems for detailed coordinate frame definitions.