Principles, roles and challenges of LiDAR for intelligent vehicles

Modern motor vehicles are undergoing major changes in response to technological and societal factors. Perception systems have become an integral part of these vehicles. In a short space of time, we have seen the use of ultrasonic sensors, now available in most vehicles, as well as the introduction of video cameras, initially as an aid to visualizing reversing maneuvers, now as an integral part of Advanced Driving Assistance Systems ( – ADAS). Images are processed using machine learning techniques to provide semantic information, including the detection and classification of different road users and their spatial position relative to the vehicle. It is also possible to extract road features such as lane markings, road signs, etc. Today, cameras are widely used in a number of ADAS functions such as automated emergency braking (AEB), lane detection, etc. Radar is used in many cases, to complement the limitations of video cameras in deducing distance information. In other words, motor vehicles have benefited from the introduction of perception sensors capable of detecting different road users or objects in their immediate environment. The information deduced is used to enhance driver vigilance or trigger emergency maneuvers, and thus improve road safety. Several applications have become standard features in motor vehicles. These perception systems form the basis of new safety-critical applications which control vehicle response, such as Adaptive Cruise Control ( – ACC) or Automated Line Keeping Assistance System ( – ALKAS), etc.

The transformation of motor vehicles is ongoing; software has become the predominant component which, combined with sensors and powerful on-board computers, is used to provide various functions improving driving comfort, fuel consumption, safety and so on. This has led to the introduction of higher levels of automation where computers progressively remove drivers from the vehicle's control loop, leading to what are now known as "autonomous vehicles" or "intelligent vehicles". In this article, we focus on a class of sensing and perception sensors known as LiDAR, an acronym for "light pulse remote sensing and ranging" (LIght Detection And Ranging), sometimes referred to as a "3D laser scanner". A LiDAR is an active sensor that works like radar, emitting pulses of infrared light instead of radio waves, and measures the time interval it takes for the signal to return after bouncing off nearby objects. The time between the output laser pulse and the reflected pulse enables LiDAR to accurately calculate the distance to each object, based on the speed of light. Its main feature is that it measures range precisely. Prior to its widespread use on intelligent vehicles, it was known as LADAR for "Laser Detection And Ranging", and was used in various mobile platforms such as...