in

To Enable Safe Autonomous Driving, Act On Reliable Measurement Data Instead Of Hunches

To Enable Safe Autonomous Driving, Act On Reliable Measurement Data Instead Of Hunches

The New York Times and FX’s recent documentary, “Elon Musk’s Crash Course,” outlines how Tesla aggressively enables and markets autonomous driving functionalities while overlooking system reliability and consumer safety for its Autopilot system.

Consumers may not fully understand the state of autonomous driving, and they may believe the technology is more advanced than it currently is. Tesla Autopilot has received significant attention, for example, but the functionality offered can be described as Level 2 advanced driver assistance systems (ADAS), where the driver still needs to pay attention while driving.

While fully autonomous or self-driving vehicles are, in my opinion, at least a decade away, there are other technologies going into our cars today that are worthy of discussion, too.

History Of Collisions

Over the past year alone, Tesla vehicles set to Autopilot have been reportedly involved in 273 crashes, making up 70% of the crashes involving ADAS since last July.

We have known for years that cars relying solely on cameras or a combination of cameras, radar and software to detect oncoming objects are not yet reliable enough. Cameras are only 2D and are easily affected by lighting conditions, which adds a layer of uncertainty to their data. While software or AI can extract more useful information from camera data, they are currently far from reliable and simply cannot make up for the data that is not collected in the first place.

According to the American Automobile Association (AAA), vehicles with current ADAS technologies still failed to consistently avoid crashes. That shows how much is yet to be done to ensure vehicle safety.

Diverse Sensor Suite For Safety

I have devoted my life to developing one such safety technology, lidar, which can work in conjunction with camera and radar to create a 3D picture of what is ahead of a car.

Lidar is a technology that scans the environment around the vehicle by sending out beams of invisible light pulses in multiple directions. The pulses collide with surrounding objects and bounce back to the receiver. Because the speed of light is constant, lidar systems create highly accurate 3D images called “point clouds,” which provide the vehicle with a constant awareness of what is happening around it, so it can react and inform drivers accordingly.

I am a firm believer that 3D perception using different modalities such as camera, radar and lidar to complement each other (just like our three senses—eyes, ears and nose) is required to correctly assess the road ahead and the obstacles on it. I believe this powerful sensor suite will be a part of every car in the future. Adding lidar to the sensor suite will effectively enhance safety because lidar has higher spatial resolutions than radar, and it works quickly to constantly generate 3D data, including the size, speed and location of a car’s surroundings even at night and in bad weather when cameras have trouble detecting.

Several lidar companies—including Luminar, Innoviz and the company I founded, Cepton—have partnered with car manufacturers to deploy lidar in the years ahead. Through this collaboration, we expect lidar to be deployed in multiple vehicle models, not just the luxury class, to benefit more consumers.

Musk has been the biggest naysayer of lidar. In a 2018 earnings call, Musk called lidar “a crutch” that will hurt the companies that invest in it. A year later, Musk called installing lidar in vehicles “a fool’s errand,” calling lidar “expensive sensors that are unnecessary.” According to The Verge, Musk walked back some of those comments last year.

Musk has missed the opportunity to lead in the technology’s implementation at scale. From a consumer perspective, because drivers tend to group all ADAS platforms together, these comments as well as Tesla’s widely reported accidents may shake consumer trust.

Manufacturer Buy-In

It might seem obvious that someone developing lidar technology would be advocating for lidar’s implementation, but those with the most at stake financially, the automakers, are also investing in lidar.

Earlier this year at an investor conference, General Motors Vice President for Global Product Development, Purchasing and Supply Chain Doug Parks pointed out “the addition of radar and lidar essentially provide a touch to the AI to enable it to rationalize and augment the camera input. It’s really important you have all three of those.” Parks continued that cars missing one of those key things “will not meet our performance and safety standards. The ones all customers expect, and the ones all customers deserve.”

The concept of a self-driving car is amazing. For his part in pushing innovation in the car industry, Musk is unquestionably brilliant and fantastically successful. But since public acceptance of ADAS is still quite low due to safety loopholes, Musk has missed the mark on preparing the consumers for the arrival of autonomous vehicles with a stronger sense of trust in the technology.

As solution providers, we should listen to the market. When drivers are asked whether carmakers should focus on building self-driving cars or improving safety technology currently in cars, 8 in 10 drivers choose the latter. There’s much for us to work on collectively. Instead of arguing over the best single technology, we need to focus on making different technologies seamlessly work together without disrupting the modern car design and cost structure. With complementing technology and redundancy, our cars can more accurately perceive their environment, thus lowering the risk of accidents.

If consumers fail to understand the current capabilities of automated systems and demand the right combination of technology in the vehicles, they may be risking their own lives. For those reasons, when it comes to the safety of our roads, automakers should act on data, not on hunches. They should focus on making cars safer now.

Source: www.forbes.com

Report

What do you think?

486 Points
Upvote Downvote

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

GIPHY App Key not set. Please check settings

Autonomous trucks will do jobs human drivers don’t like, panel says

Autonomous trucks will do jobs human drivers don’t like, panel says

(video) Advanced perception training: child pedestrians detection