Automotive LIDAR systems: the design engineer’s guide

The automotive world is going through a period of unprecedented change as vehicles move in stages from the entirely driver-controlled cars we are familiar with, through those that help us drive to the eventual goal of vehicles that can drive themselves in all situations and conditions.

This path is very well defined and the American Society of Automotive Engineers (SAE) has developed a model so that automakers are all speaking the same language with regard to the capabilities of their vehicles.

The lowest level of automation is Level 0 where there’s no automation at all and everything relies on the driver. Most of the vehicles on the road fall into this category. Levels 1 and 2 include some assistance features that are often collectively referred to as Advanced Driver Assistance Systems (ADAS). Included in this are features such as automated emergency braking (AEB), lane keeping assist (LKA), adaptive cruise control (ACC) and stability control. The SAE distinguishes Level 1 as being where one of these systems operates in isolation and Level 2 is where two or more can operate together.

While deployment is in the early stages, it’s expected to grow rapidly as many regulators are mandating these systems for any vehicle that wants to achieve a 4- or 5-star safety rating.

Level 3 is the start of full autonomy and here vehicles can operate autonomously in simpler environments, such as on highways where there are no pedestrians, no complex intersections and all the vehicles are moving in the same direction at similar speeds. The driver is essentially a ‘caretaker’, monitoring the progress of the vehicle and taking over immediately if the vehicle requires assistance.

Full autonomy is covered by Levels 4 and 5. These vehicles are capable of autonomous operation in all conditions and the driver is solely a passenger. While Level 4 vehicles will retain controls such as pedals and a steering wheel, Level 5 will not.

Technologies to support autonomous driving

There is a significant step between Levels 2 and 3 in terms of the technology required to operate safely and effectively. Absolutely critical is the ability to identify obstacles in the path of the vehicle, including fixed and mobile obstacles such as animals or pedestrians. A number of technologies are being developed for this including vision sensors, LIDAR and radar. While each can work independently, they each have different advantages and disadvantages and are often combined via sensor fusion to provide the accuracy needed for safety-critical systems.

Deploying these essential detection systems puts a significant strain on the infrastructure of the vehicle, especially as more accurate (and therefore safer) systems require higher resolution sensors with greater frame rates. This leads to much higher levels of data within the vehicle which older networking technologies such as CAN struggle to keep pace with.

A special version of Ethernet that does away with the random access CSMA/CD and provides deterministic / time sensitive operation has been developed for use in vehicles. Able to handle data rates into the gigabit region, automotive Ethernet is able to prioritise safety-critical traffic over more mundane information.

An overview of LIDAR

LIDAR is sometimes referred to as ‘laser radar’ and it’s a time-of-flight (ToF) system that’s able to both detect an object and calculate its distance. Although relatively new to vehicles, LIDAR is already used in 3D aerial mapping and smart weapons, among other applications.

LIDAR works by sending out a beam of light which will bounce off any objects within range and a part of the signal will return to the LIDAR unit and be detected. By knowing the time taken for the light to make the round trip (the ‘time of flight’) and the speed of light, the distance between the sensor and the object can be calculated.

Figure 1: LIDAR bounces laser light onto objects to determine their location

Alternative technologies such as ultrasonic sensors can perform a similar function but, as ultrasonic waves are heavily attenuated in air, they are really only suited to short-range applications such as parking sensors. Vision sensors require a significant amount of processing to deliver useful data, so they’re an expensive solution although they do have the benefit of being the only sensor type that can detect colour, which is a requirement for some applications.

Radar and LIDAR are quite similar in principle, although radar uses radio waves and LIDAR uses laser light. Both are able to work over distances as large as 200 metres, although for really close work radar has an advantage. As LIDAR has a very small wavelength, it’s much better than radar in performing 3D characterisation of objects with small features and LIDAR also has a much better field-of-view (FOV) than radar.

Although radar generally performs better than LIDAR in adverse weather, using light with a wavelength around 1,550 nm does significantly improve LIDAR performance in bad weather. In common with vision sensors, LIDAR can be affected by lighting conditions.

Until now, the cost of LIDAR has made it prohibitive, although costs are dropping dramatically. The US$50,000 price point has reduced to around US$10,000 and, as the popularity grows, some experts are forecasting a sub-US$200 price within a couple of years.

Types of LIDAR

Earlier LIDAR implementations were mechanical, meaning they used a rotating assembly and high-grade optics to create a solution with a 360-degree FOV. These were typically the ‘domes’ that were seen on experimental vehicles and, although they were bulky and expensive, they did offer an excellent signal-to-noise ratio (SNR).

Solid state LIDAR has no rotating components and, as a result, has a much reduced FOV. However, the substantially lower cost means that a sensor can be placed on the front, back and each side of a vehicle. The data from the four sensors is then fused to give a LIDAR signal that very closely approximates the FOV of a more expensive mechanical system.

There are several types of solid-state LIDAR currently available. MEMs-based systems use micro-miniature mirrors to direct the beam, but the systems are fragile and require precise alignment of multiple mirrors which, along with the reduced operating temperature range, limit their suitability for automotive applications.

Flash LIDAR works much like the flash on a camera, sending out a single large-area laser pulse capturing the entire reflected scene in a single image, which makes it much faster than the mechanical systems. As it is instantaneous, flash LIDAR is much less affected by vibration, although the high peak laser power needed can be a design challenge.

Optical Phase Array (OPA) LIDAR works in a similar way to phased-array radar with multiple beams that are delayed by increasing amounts.

Frequency-modulated continuous wave (FMCW) LIDAR moves away from the ToF principle and uses brief chirps of frequency-modulated laser light. By measuring the frequency and phase of the returned light, the system can determine both distance and velocity. While the optics needed and computational power required are advantages of this method, generating the chirp is potentially challenging.

The passive and connectivity challenge

Figure 2: Overview of a typical LIDAR implementation

As can be seen in the block diagram, LIDAR systems are relatively complex with multiple sub-systems that encompass both digital and analogue electronics. While semiconductors have an important role to play, passive components and interconnect between the sub-systems are equally important, especially in areas such as noise suppression which is essential to the operation of automotive LIDAR systems.

Much of the automotive challenge relates to the physical environment associated with automotive applications. It is often at an elevated temperature and high levels of vibration and shock put even more demands on components.

Modern components are available to meet the demands of automotive applications, and an increasing number meet the needs of AEC-Q200, allowing designers to specify them with confidence. Examples include metal composite type power choke coils that are able to operate at temperatures up to 150C and withstand high levels of continuous vibration. The shielded construction supports their use in many applications where EMI is a consideration, such as DC/DC conversion for automotive LIDAR systems.

Summary

LIDAR plays an essential role in giving vehicles the spatial awareness they need to achieve autonomous operation. LIDAR’s principle of operation is similar to radar, but it uses laser light and ToF principles to generate a 3D map of the vehicle’s surroundings. Although relatively expensive, costs are plummeting and are expected to reach a point very soon where LIDAR becomes a viable technology for almost all vehicles.

Solid-state LIDAR seems to offer the best solution for mainstream vehicles, as multiple sensors can be placed around the vehicle and have their data combined to give a FOV that approximates the more cumbersome and expensive mechanical systems.

However, critical to any automotive system are the passive components and interconnect devices that are essential to developing complex systems such as automotive LIDAR. Recent advances in technology are delivering components that are qualified for long term use in the highly demanding automotive environment.

Below we’ve highlighted our leading suppliers for components suited to automotive LIDAR systems.

If you require advice on selecting the right components for your design, our technical specialists are on hand to help.