I recently joined several industry colleagues at Analog Garage, ADI’s technology incubator, to discuss the continued evolution of autonomous transportation. This is a topic we’re all immersed in daily and realize is shaped by several obstacles – design, regulatory, legislative, and technology challenges included. While all these influence how driverless cars are hitting the road, they also prompt one of the biggest questions we currently face:
Despite the lack of a concrete answer, it’s still an exciting time for autonomous transportation given the amount of innovation we’ve seen in recent years. Much of this invention is coming out of Boston, including engineers working on the technology out of Analog Garage. That’s where I was joined for a discussion on the future of autonomous transportation by several experts, including Eryk Nice, vice president of autonomous systems at nuTonomy, a division of Aptiv’s Automated Mobility Group, Anita Kim of the U.S. Department of Transportation’s Volpe Center, and Sanjay Aggarwal of VC fund F-Prime Capital.
This small group concurred that if we dive deeper into autonomous driving, we see that much of the foundational sensor technology is already used in cars today. Cameras are ubiquitous, and RADAR technology is starting to be deployed widely in applications such as automatic emergency braking. Unfortunately, this sensor technology is a long way from where we need to be to attain the higher levels of autonomy that will truly enable self-driving vehicles.
Consider that there’s no current LIDAR system in today’s cars, despite the fact this specific sensing technology is receiving huge industry investment. Finding a way to implement LIDAR while adhering to size, weight, power and cost budgets needs to be top of mind. Additionally, for the industry to successfully shift from a warning to an actuation system, there’s more that must be done from a performance perspective to ensure a much lower false positive rate. Today, these perception systems exist to warn users that an object of interest is in their vehicle’s region of operation, but too often the performance of these sensors is not sufficient to form a fundamental input to actuate the vehicle or change what the car is doing.
Autonomous vehicles – either in simulation or on the road – have not come close to approaching that number. To meet minimum safety thresholds, self-driving cars must continue to be tested in adverse weather and road conditions. For example, Boston has different – and more challenging – driving conditions than Las Vegas. If driverless cars can safely navigate Boston traffic, interpret road signs that confound many natives, and avoid the inevitable springtime potholes, that will go a long way toward demonstrating the maturity of the technology.
Also, in looking at cars today operating with some level of semi-autonomy, the performance of these systems varies widely, which for consumers can raise issues of trust. If we get into one car and it operates and actuates one way, and then we travel in another car that operates differently, how will consumers react to that experience? If we can standardize safety systems across different vehicles from different manufacturers, then the technology can be delivered to the market more consistently – and broadly. Today, however, that lack of uniformity is apparent in how automakers are developing and deploying safety systems – an area the industry must address.
Vehicle industrial design is another key opportunity where consumer trust can be fostered and enhanced. Robo-taxis will be the first embodiment of autonomous vehicles. In these applications, sensors will be prominently positioned to show the occupant that the system has the eyes to “see” around the vehicle. This will serve to reinforce to the public that the core sensor and computing frameworks are up to the challenge to operate the vehicle autonomously. This, in turn, will feed into mass-market automotive applications as the technology matures and customers demand (and trust) more advanced autonomous features. In these applications, the sensors are “under the hood” and the user must trust they will operate well under all circumstances.
I want to thank Eryk, Anita, and Sanjay for joining this discussion at Analog Garage in Boston. Those working in this nascent industry have a unique opportunity in front of them. It is rare when we can collectively address some of the most difficult engineering challenges of our time. However, it’s important to realize these have broad positive social impact by reducing the impact of transportation on the environment, improving productivity, and – most importantly – saving lives.
Besides implementation, safety is a challenge as well.