Driverless Startup Wayve Wants to Make LIDAR Obsolete

author: Michael Cheng   

LIDAR is viewed as a crucial component in autonomous vehicles. Capable of processing accurate images of surroundings, the technology increases reliability and safety during driverless operation.

Is it possible to build a self-driving car without LIDAR? UK-based startup Wayve thinks so.

The company believes end-to-end machine learning, cameras and satellite navigation are enough to power the autonomous functions of self-driving cars. Capable of adapting to new environments quickly, the startup's somewhat controversial approach has received mixed feedback from other developers and auto companies.

No LIDAR, No Problem?

Wayve's autonomous platform relies on imitation and reinforcement learning to enhance driving capabilities. Uncertainty is baked into the models, so that deep learning is maximized based on input data. Computer vision is used for guidance along various routes. Such methods do not require expensive HD-maps, hand-coded rules and automotive safety sensors. According to the startup, such solutions also allow autonomous vehicles to maneuver through ‘never-before-seen' roads.

"With each safety-driver intervention, our system learns and will improve, rather than buckle with scale. It will take us longer to reach our first deployment, but we are riding a fundamentally different curve," said the startup.

The benefits of Wayve's approach to autonomous driving are numerous. The computing power of a laptop computer is all that is required to support the platform's features. As a result, less computing power is needed inside the vehicle (compared to traditional self-driving systems).

Moreover, with the absence of bulky sensors and massive computing equipment, cost is greatly reduced. Wayve cited in a recent blog post that its sensor and computing costs are only 10 percent of costs associated with current autonomous platforms that utilize LIDAR and HD-maps.

Interestingly, the startup's system is usable in bad weather conditions, including rain and snow. It also does not require endless hours of training or virtual simulation to learn proper driving behavior. Videos of the driverless platform in action shows a vehicle maneuvering (at slow speeds) through tight streets, intersections, traffic and around cyclists.  

Redundant Safety

The case for building autonomous vehicle platforms without LIDAR moves away from implementing redundant safety features. Such safety protocols, which can be found in aircraft hydraulic systems and large-scale data servers, exist because they work very well. Although the equipment used to implement redundant safety may not be utilized on a regular basis, it can prevent the onset of critical failures or major accidents.

"The whole point of self-driving cars is to be safer than a human driver," explained Matt Weed, Director of Technology Strategy at Luminar Technologies.

"Why you would eliminate technology that maximizes safety inputs doesn't compute? You want to be able to get as much good information about the world as you can."

Wayve's self-driving solutions are still under heavy development; therefore, it is still too early to tell how reliable they would be in real-world scenarios, on busy roads and highways. If the startup can prove the effectiveness of its methods (through extensive testing and public trials), then it would certainly be game-changing for the driverless sector.

Michael Cheng

Michael Cheng is a legal editor and technical writer with publications for Blackberry ISHN Magazine Houzz and Payment Week. He specializes in technology business and digesting hard data. Outside of work Michael likes to train for marathons spend time with his daughter and explore new places.