The issues automakers face with LiDAR and autonomous driving are similar to what geospatial professionals see in unmanned aerial vehicles (UAVs) and even survey technology. As automakers address their challenges with LiDAR, other users of LiDAR technology are likely to benefit.
Starting Down the Road
The price of an average new car is up 29 percent over the past decade.
Some consumers, especially millennials, are now opting to lease their vehicles, but even that is getting expensive.
For many consumers, the best solution appears to be buying used vehicles.
This impacts the automotive industry, which wants to sell new inventory and not be left with depreciating assets, like used vehicles or vehicles coming off lease that must be depreciated.
To answer the challenge of providing new, innovative vehicles for consumers that will attract wallet share, auto makers are incorporating more technology into cars, but these technologies have also contributed to increases in prices.
Higher cost is one of the challenges LiDAR technology faces as it’s incorporated into autonomous vehicles. It also faces challenges in areas like size and reliability. But before we investigate the challenges for LiDAR, it’s important to understand why LiDAR plays such a key role in the operation of vehicles.
LiDAR’s role in autonomous vehicles
Without the benefit of LiDAR, autonomous vehicles would be driving blind.
Whether LiDAR is mounted as a rotating device on top of a vehicle, or on the hood, LiDAR becomes the “eyes” of the vehicle and provides a 360-degree view of a traffic scene. LiDAR sends out thousands of laser pulses every second that bounce off of surrounding objects such as other automobiles that are in the vicinity of the vehicle. The pulses are then translated into light reflections.
The light reflections create a 3D point cloud and the autonomous vehicle’s onboard computer system transforms this information into a 3D representation that captures the speed of the light and the distance that it covers. The end objective is for the vehicle to determine its position in relation to the position of other surrounding objects.
Down the line, the vehicle’s onboard computer system has the potential to further operationalize this data—such as commanding the brakes to stop or slow a vehicle.
|Terrestrial Mobile LiDAR||38%|
|Time-of-Flight Terrestrial Stationary Laser Scanners||34%|
|Phase-Based Terrestrial Stationary Laser Scanners||33%|
Source: 2018 Survey & Mapping Deep Dive: Laser Scanning & GIS CLEAReport
LiDAR usage is on the rise among surveyors and geospatial professionals. In the 12 months prior to the Survey & Mapping Deep Dive study on laser scanning, most respondents had purchased laser scanners or LiDAR. Over 70 percent planned to make future purchases as well.
The challenges for LiDAR
As an autonomous vehicle’s “seeing eye,” LiDAR plays an indispensable role in self-driving cars. However, there are also inherent challenges in LiDAR that must be overcome.
Here are the core challenges, and what the industry is doing about them:
Cost - In 2017, the Los Angeles Times reported that the cost of LiDAR per car was around $75,000. In 2019, we are now hearing that the cost of LiDAR has fallen to around $1,000 per car. Apple, which has entered the autonomous vehicle space, is actively seeking low-cost, mass produced LiDAR that can be manufactured by using standard semiconductor technology — potentially lowering the cost of each vehicle LiDAR unit into the hundreds of dollars.
This still adds significant cost onto a car, but it is a 75:1 reduction from what it was only two years ago.
Size - Apple is working on reducing the size of LiDAR units and companies like Luminar are building LiDAR that draws only 15 watts of power but has an acceptable range of 250 feet.
Both are steps in the right direction, but will reducing LiDAR’s form factor create other risks such as overheating of the smaller units, or less ability to capture high resolution images?
These are some of the tradeoffs currently being weighed as efforts continue to reduce overall LiDAR form factors.
Reliability - As LiDAR units get smaller, new calculations must be made to adjust for factors such as heat build-up in sensors that can compromise LiDAR performance and also the accuracy of the 360-degree vision that the technology is supposed to deliver.
There are also environmental factors that LiDAR technology must be able to cope with – such as not being able to produce accurate results in very dark settings, or in fog or snowstorms — and being able to function if the unit itself gets dirty and isn’t properly maintained.
Work in these areas is ongoing because LiDAR, like any technology, is going to have its limits as well as its advantages.
One idea from the VTT Technical Research Center in Finland was the deployment of LiDAR filtering technology for processing environmental perception data in snowy conditions, and also teaming LiDAR with radar for better results in snow. LiDAR sensor suppliers in the automotive industry are also working on ways that these sensors can “self clean” so they can continue to function optimally.
Integration - Integrating LiDAR with hardware on vehicles and also with other systems such as radar, GNSS, electrical systems and vehicle and cloud-based software infrastructure is a daunting challenge.
In 2019, Jake Li, the business development manager for automotive LiDAR for Hamamatsu, a Japanese photonics component supplier, discussed the optical challenges of LiDAR in self driving cars.
“If you want object classification, you need to know that you’re seeing what the object is, at close range or long range,” said Li, who added that LiDAR’s 0.1 degree angular resolution “basically allows that to happen,” in contrast to systems like radar.
Li also pointed out that LiDAR had advantages over camera systems, which have trouble in low-lighting and variable-lighting conditions that can temporarily “blind” the camera. Unfortunately, LiDAR also shared one key disadvantage with cameras: not functioning well in conditions of adverse weather.
These are areas where hardware and software can be improved—but improvement and performance gains also require integration and teaming of LiDAR with other technologies, all orchestrated by a single software infrastructure that is smart enough to run vehicles, blend technologies to respond proactively to different conditions, and communicate with point clouds.
Vendors like Siemens are taking on the challenge of creating an intelligent infrastructure that integrates the functionality of LiDAR, radar, point clouds and other technologies to handle the complexities of driving.
Like other technologies, LiDAR is still a work in progress in autonomous vehicles.
However, new approaches that promise to make LiDAR smaller, lighter and cheaper will continue to reinforce its central role in autonomous driving. So, too, will new technology breakthroughs such as frequency-modulated continuous-wave (FMCW) systems, which extract time and velocity information from the frequency shift of returning light, and can produce richer streams of data; or solid-state LiDAR that doesn’t use beam scanning at all.
While there are the detractors such as Tesla’s Elon Musk, who prefers cameras over LiDAR, the new technology breakthroughs for LiDAR have not gone unnoticed by other automakers (or the geospatial community).
In 2019, BMW announced it is going with a solid-state LiDAR system for the company’s self-driving vehicles, which it plans to put into production by 2021. Audi also committed to LiDAR for its self-driving vehicles. The two automakers join Ford, GM Cruise, Uber and Waymo in LiDAR adoption.