LiDAR & Radar in AI & Artificial Intelligence

LiDAR works by firing thousands of laser pulses per second and measuring the 'Time of Flight' (how long it takes for the light to bounce back). This results in a Point Cloud—a highly accurate 3D representation of every surface the laser hits. For a robot, this is superior to a camera because it provides 'Metric' data (e.g., 'that wall is exactly 4.23 meters away'). However, processing millions of points in real-time requires efficient data structures like Kd-Trees or Octrees to maintain a high cycle frequency.

Radar uses radio waves rather than light. While it has lower resolution (it can't see the 'shape' of a pedestrian easily), it has two massive advantages. First, radio waves penetrate fog, snow, and heavy rain, allowing the robot to 'see' when cameras and LiDAR are blinded. Second, it uses the Doppler Effect to measure the relative velocity of moving objects in a single frame. This makes it the primary sensor for Adaptive Cruise Control and Collision Avoidance systems.

Raw sensor data is messy. A LiDAR point cloud includes the ground, which isn't an obstacle, just the surface the robot is driving on. We apply Ground Filtering algorithms (like RANSAC plane fitting) to identify the road and remove it from the obstacle list. We then perform Sensor Fusion, using the high-resolution LiDAR to detect 'What' and 'Where', and the Radar to detect 'How Fast'. This combined model is much more robust than any single sensor could ever be.

Cameras see colors, but LiDAR and Radar see 'Depth' and 'Velocity'. These sensors are the eyes of autonomous vehicles, allowing them to map the world in 360 degrees, even in total darkness.

LiDAR (Light Detection and Ranging) uses laser pulses to create a 'Point Cloud'. Each point represents a physical reflection in 3D space.

Radar uses radio waves. It's lower resolution than LiDAR, but it can see through rain and fog, and it can measure the 'Velocity' of other cars instantly.

Checkpoint: What is a 'Point Cloud' in the context of LiDAR?

• →A type of weather system
• →A set of data points in space, representing the external surface of objects as detected by lasers

We use 'Ground Filtering' to remove the floor from the LiDAR data, leaving only the obstacles (cars, pedestrians, trees) for the robot to avoid.

By mastering LiDAR and Radar processing, you learn to interpret the raw energy of the world, turning light and radio waves into a safe path forward.

Checkpoint: Why do we use BOTH LiDAR and Radar on autonomous cars?

• →To make the car more expensive
• →Redundancy: LiDAR is high-resolution, but Radar works in bad weather (fog/rain) where LiDAR fails

Sensor processing mastered! You've learned to see in 3D. Ready to build maps while you drive with SLAM?

Compute a Real Time-of-Flight Distance. Finish computing distance from a LiDAR pulse's round-trip travel time.