The Sensor Zoo
Vetted specifications for the cameras, LiDARs, IMUs, and safety scanners used in the physical AI examples
The Sensor Zoo lists every sensor in the mlsysim.Sensors registry. Each entry carries units and a provenance record, so the timing and bandwidth numbers that feed information-age and ingestion budgets trace back to a datasheet.
Use these sensors to budget the time between an event and the moment the system can act on it: exposure, readout, transport, and processing. Load any sensor directly in Python, for example camera = mlsysim.Sensors.Camera.Sony_IMX477.
Cameras
| Sensor | Resolution | Frame rate | Shutter | Exposure | Readout | Interface | Power |
|---|---|---|---|---|---|---|---|
| Intel RealSense D435i | 1920 × 1080 | 30 Hz | rolling | — | — | USB 3.1 Gen 1 | 2.0 W |
| Sony IMX296 | 1440 × 1080 | 60 Hz | global | 5.0 ms | 5.0 ms | MIPI CSI-2 (1-lane) | 0.3 W |
| Sony IMX477 | 4056 × 3040 | 60 Hz | rolling | 16.0 ms | 16.6 ms | MIPI CSI-2 (4-lane, 10-bit full-resolution 60 fps mode) | 0.6 W |
LiDAR
| Sensor | Channels | Scan rate | Max range | Range accuracy | Interface | Power |
|---|---|---|---|---|---|---|
| Ouster OS1-64 | 64 | 20 Hz | 120 m | 25 mm | Gigabit Ethernet (UDP) | 18.0 W |
| Velodyne HDL-64E | 64 | 10 Hz | 120 m | — | Ethernet (UDP) | — |
Inertial measurement units
| Sensor | DoF | Sample rate | Gyro noise density | Accel noise density | Interface |
|---|---|---|---|---|---|
| Bosch BMI088 | 6 | 1,000 Hz | 0.014 deg/Hz⁰⋅⁵/s | 0.00016 g_0/Hz⁰⋅⁵ | SPI / I2C |
Safety laser scanners
| Sensor | Protective field | Response time | Scan angle | IEC 61496 | Performance level |
|---|---|---|---|---|---|
| Keyence SZ-V32 | 5.7 m | 80 ms | 190 deg | Type 3 | PL d, Category 3 |
| SICK nanoScan3 Core (NANS3-CAAZ30AA1) | 3.0 m | 85 ms | 275 deg | Type 3 | PL d |
Where these numbers are used
Sensor timing sets the floor on how stale an observation is by the time the system acts on it. The Volume IV chapters use these entries for information-age budgets (half the exposure plus readout, transport, and processing) and for multi-camera ingestion bandwidth. The underlying formulas are in physics.robotics.perception.