Luxonis OAK-D Lite
Stereo depth plus a 4 TOPS vision processor that runs your neural network inside the camera. Depth-aligned detections come out of the USB cable already solved.
The role it plays
Sensors split into proprioception — encoders, IMUs, joint torque, motor current — which tells the robot about itself, and exteroception — LiDAR, depth, force, GNSS — which tells it about the world. Build proprioception first. A robot that does not know where its own joints are cannot use a good world model anyway.
Typical use: Object detection without a GPU host, Pi-based robots, person tracking.
What the numbers actually mean
The headline figure is 0.2 – 12 m (ideal 0.4 – 3 m) (depth range), with 4 TOPS (Intel Movidius RVC2) as the next most load-bearing number (ai compute). At $150 in the sensors tier it sits in the mid range: enough money that you should plan the mounting and wiring before ordering, but not so much that one mistake ends the project.
How it compares
Same subsystem, closest alternatives. The primary spec column is usually the one that decides it.
| Part | Price | Primary spec | Second spec |
|---|---|---|---|
| Luxonis OAK-D LiteDepth camera with on-board AI | $150 | 0.2 – 12 m (ideal 0.4 – 3 m) | 4 TOPS (Intel Movidius RVC2) |
| AS5600 magnetic encoderAbsolute rotary encoder | $3 | 12-bit (4096 counts/rev) | I²C (also analog / PWM out) |
| AS5047P magnetic encoderAbsolute rotary encoder | $10 | 14-bit (16384 counts/rev) | SPI + ABI quadrature + UVW |
| CUI AMT102-VCapacitive incremental encoder | $25 | up to 2048 PPR (configurable) | Quadrature A/B/Z |
| BNO085 9-DOF IMUIMU with onboard fusion | $25 | 3-axis accel + gyro + magnetometer | Quaternion / Euler / rotation vector, plus raw |
Where it shows up
Platforms in this atlas that use something in this class: TurtleBot 4.