Time-of-Flight (ToF) Sensors

Stereo vision relies on two cameras and heavy computational processing to triangulate depth by matching features in images, which can fail in low-texture environments (like white walls). ToF is an active sensor that projects its own light, requiring less CPU power and working perfectly on texture-less surfaces and in complete darkness.

LiDAR typically offers longer range (100m+) and high outdoor reliability but is often expensive and mechanically complex (if rotating). ToF cameras are solid-state, cheaper, and provide dense 3D images (thousands of points) instantly, making them superior for close-range interaction, obstacle avoidance, and gesture recognition, though usually with a shorter range (up to 10-20m).

Historically, yes, as the sun emits high levels of infrared light that can wash out the sensor. However, modern AGV-grade ToF sensors use narrow bandpass filters and distinct modulation frequencies to subtract ambient light, allowing them to function reliably in outdoor environments or near windows.

Multipath interference occurs when light bounces off multiple surfaces (like a corner or a shiny floor) before returning to the sensor, causing distance errors. Advanced ToF algorithms filter these outliers by analyzing signal confidence and shape, though positioning sensors to avoid highly reflective corners is a best practice.

Dark objects absorb infrared light, reducing the signal strength returning to the sensor. While modern ToF sensors have high dynamic range to detect low-reflectivity items (as low as 10% reflectivity), the effective range for black objects is typically shorter than for white objects.

Indirect ToF (iToF) is generally effective up to 5-10 meters, offering high precision for indoor navigation. Direct ToF (dToF) sensors, often found in newer automotive and industrial applications, can reach ranges of 20-50 meters or more, bridging the gap between cameras and LiDAR.

Most industrial ToF modules come factory-calibrated for lens distortion and temperature compensation. However, extrinsic calibration (determining the sensor's position relative to the robot's center) is required during integration to ensure the 3D data aligns correctly with the robot's map.

In Indirect ToF, multiple sub-frames are captured to calculate phase shift. If the robot moves fast during this capture, artifacts occur. High-performance global shutter sensors minimize this, and dToF (which uses single pulses) is largely immune to motion artifacts.

Absolutely. Most major ToF manufacturers provide ROS 1 and ROS 2 drivers. The output is typically published as a `sensor_msgs/PointCloud2` or `sensor_msgs/Image` (depth), which can be directly ingested by navigation stacks like Nav2 for obstacle marking.

Industrial ToF sensors are designed to be Class 1 Laser Products, meaning they are safe under all conditions of normal use. They typically use VCSEL arrays that diffuse power over a wide area, and built-in hardware watchdogs shut down the emitter if a failure is detected.

ToF sensors generally consume between 1W to 5W depending on the illumination power required for the desired range. This is significantly lower than spinning LiDARs (which often pull 10W+) but slightly higher than passive stereo cameras, making them very efficient for battery-powered AGVs.

Generally, no. ToF sensors rely on reflection. Transparent materials like glass or acrylic usually allow the light to pass through (measuring the wall behind it) or cause specular reflections that result in noisy data. Ultrasonic sensors are often used in tandem to detect glass surfaces.