New Eyes, New Ears, New Brains: Inside the Next Generation Argo AI Self-Driving System
Toggle Filter Camera
Safe and reliable: Those are the watchwords for everything we do at Argo. We are committed to building a self-driving system that, above all else, makes the roads safer for everyone who uses them. But we’re also dedicated to making self-driving work as a business.
As such, every sensor, circuit board, and wire harness that goes into our partners’ autonomous vehicles needs to be manufacturable at scale. They need to be cost-effective. And they need to be automotive grade — meaning they perform consistently at all times of day and in a wide range of weather conditions — from the salty humidity of Miami to the wintry cold of Detroit.
The launch of our fourth generation self-driving system (SDS) on our fleet of Ford Escape Hybrid self-driving test vehicles represents a milestone on this phased journey to safe autonomous driving by positioning us squarely in the productization stage on the way to commercial deployment. It comprises “product intent” components, which mean that they have the functional performance we need, can be manufactured affordably, and offer automotive-grade ruggedness, reliability, and durability.
Our new SDS leverages customized components — not off-the-shelf stuff — including high-resolution cameras, lidar, radar, microphones, and inertial sensors, that meet rigorous industry safety standards. This is important for a few reasons. First, it puts Argo in the position to work with our partners at Ford and Volkswagen to transition into the component validation stage to ensure they all work together in advance of production. Second, we believe it establishes Argo as the first self-driving technology company to develop automotive-grade components — hardware built to withstand extreme thermal conditions and maintain structural integrity over time — that can be manufactured in significant volumes.
Our fourth generation SDS is loaded with sensors and computers to enable autonomous operation. These serve as the eyes, ears, and brain of the vehicle and allow it to detect objects both static and dynamic; assess the speed and trajectory of vehicles and other road users in the vicinity; and enable it to react to ever-changing road dynamics. For this generation, Argo developed the custom specifications for each component working with suppliers and our in-house engineers to test and verify that everything meets the needs of our system:
- Lidar: The new SDS boasts lidar sensors that enable the system to better see where it is in relation to other vehicles, road users, and infrastructure. The new long-range lidar is now a single integrated unit, allowing for easier maintenance, improved thermal controls, and denser coverage at longer distances. The new lidar base contains water jets for cleaning and fans for cooling, allowing the sensors to efficiently operate in extreme temperatures and for the optical windows to be automatically cleaned if they’re ever obstructed by rain or dirt.
- Radar: Another big advancement is in the radar devices. Our new radars nearly double their range of detection, an improvement which, when coupled with the expanded detection capabilities of the long-range lidar, allows our vehicles to travel safely at higher speeds.
- Cameras and Image Processing: The new SDS incorporates a host of high-resolution far-field and near-field cameras loaded with new custom image sensors that offer more advanced pixel technology. This improved vision capability is crucial in high dynamic range scenes, for instance when a vehicle is emerging out of a dark tunnel and into bright sunlight. Now, the camera’s pixel technology combined with a custom image processing pipeline can handle challenging lighting scenarios, providing accurate perception of the world without loss of visibility.
- Audio Detection: The addition of three microphones allows our SDS to effectively listen for emergency responder vehicles and react accordingly as they approach, even when those vehicles are not visible to cameras, lidar, or radar.
- On-Board Computers: The SDS embraces a redundant computer design in which we have two independent computing systems that serve to maintain safe operations. The main computer, called the Autonomous Vehicle System (AVS), controls all of the self-driving software. Processing data coming from the sensors that constantly scan the road ahead and all around the vehicle, the AVS is what drives the vehicle, including performing emergency or evasive maneuvers, if necessary. There is also a separate computer called the Complementary Autonomous Vehicle System (CAVS). Running in parallel with the AVS and always active, the CAVS is designed to perform collision-mitigation functions if required, and to offer critical backup in case of a main system failure.
A focus on safety
Building from the lessons of the first three generations, our efforts to build this generation self-driving system continued to refine our methods to design, model, and evaluate both system and component performance over time. Through Failure Modes and Effects Analysis, or FMEA, we worked with Ford to diagnose hardware issues and identify all the ways in which an individual module might fail. We identified mechanisms to assess the severity or risk of each failure.
From there, we collaborated with our partner to improve performance overall, and built diagnostic checks into the SDS to identify, isolate, and, in some cases, solve problems on the fly. We also teamed up with chip manufacturers to make sure that the hardware we install in our on-board computers is optimized for use in self-driving vehicles that will operate in a wide range of different road conditions. That means both computers are built to automotive grade standards and have been strengthened to withstand the rigors of operating in an urban environment, such as the bumps and jolts that come with potholed roads. Another aspect of safety and digital security is the fact that the computers use different detection algorithms so the backup computer has a unique perception ability which improves the robustness of response in an unexpected situation. The computing systems also have independent communications pathways and power sources which serve to retain their operational independence.
We also addressed reliability, since our self-driving cars have so many software systems on-board. Our new SDS comprises dedicated features and processes that monitor for faults in hardware and measure the behavior and performance of mission-critical operations. In addition, we conducted extensive reliability and environmental testing in several of our test cities across the United States, with tests engineered specifically to the climate and particularities of each market. For example, we ran the new hardware through effects of salt spray to make sure it did not fail in the coastal conditions of Miami. We also ran the SDS through extreme temperature conditions that mirrored the cold of a Midwestern winter in Pittsburgh or Detroit, as well as the heat of an Austin summer.
Finally, we focused on internal temperature management, installing fans and cooling systems to make sure our sensors and computers do not overheat even during extreme daytime temperatures. The results are increased safety and expanded performance — on both the hottest and the coldest days of the year.
All of these efforts exemplify Argo’s commitment to testing to ensure our SDS performs optimally in a variety of driving conditions in different cities. We believe this advancement moves Argo another step closer to its goal of making a smarter and safer self-driving system that is both reliable and enjoyable for people in cities all over the world.