National

AV companies are largely developing their cars as a package deal, which requires extraordinary expertise and investment in vehicle technologies, software and cloud-based systems.

Why it matters: This strategy could ultimately take a toll on innovation and competition by limiting the possibility of interchangeable components that could drive down costs and bring AVs to market faster.

The big picture: In the vertically integrated approach, AV companies are developing all of their tech stack either in house or within proprietary partnerships. This means that an AV's components will be tied to just one company's ecosystem, even though it's unlikely any one company will dominate the field across vehicle, software and cloud.

This system results in better integration of hardware, software and cloud-based systems.

It also yields greater control over customer experience and software updates, and makes contracting out cybersecurity systems easier.

The expertise needed to excel at all three components is expensive and time-consuming, even for tech giants like Google and Amazon. This could delay when AVs go to market, and make them more expensive for fleet owners and end users. 

It could also impede innovation and competition, because AV companies would not have the flexibility to change systems or partners if better tech becomes available.

What's happening: Most AV companies don't have Google's budget and deep bench of multifaceted engineering expertise, and instead are pursuing proprietary partnerships. 

For example, Ford has recently invested $1 billion into Argo AI and Amazon recently invested in Aurora in its efforts to edge into the AV space.

As an alternative, AV companies could strategically collaborate with tech suppliers across the industry to make the most promising versions of any tech run in all AVs.

This would also be cheaper than building entire teams in house, and would allow for greater flexibility as tech companies compete.

Even Ford eventually abandoned Henry Ford's vertically integrated vision in favor of outsourcing technology to the best suppliers.

The bottom line: If AV tech is designed to be interchangeable, the market could be more competitive, and AVs could be available sooner and at a lower cost.

Bibhrajit Halder is the CEO of an early-stage AV startup and has worked on autonomous vehicles at Ford, Caterpillar and Apple. He is also a member of GLG, a platform connecting businesses with industry experts.

Too much proprietary tech could hurt the AV market

Exclusive partnerships could stifle innovation and delay deployment.

Illustration of a computer wearing a seat belt.

Transportation

Autonomous vehicle technology

Expert Voices

Technology

Business

AV companies are largely developing their cars as a package deal, which requires extraordinary expertise and investment in vehicle technologies, software and cloud-based systems.Why it matters: This strategy could ultimately take a toll on innovation and competition by limiting the possibility of interchangeable components that could drive down costs and bring AVs to market faster.

From the Valley to D.C., Big Tech players like Facebook, Google and Amazon are under more scrutiny than ever as new technology develops and privacy and antitrust concerns grow in lockstep with companies’ ambitions.

Instead of trying to design fully automated vehicles from scratch, the way forward might be to adapt limited technologies that have already proven themselves in specific settings, and gradually add capability.

Why it matters: Adapting new technology from successful use-cases where safety is paramount, like in mining, could assuage public concerns about AVs, and accelerate AV development and production. 

How it works: AV development and production could be rooted in successful deployment in constrained environments and then scaled up from there. 

For example, AVs are already used in limited defense applications, in mines, and in factories. 

Next they could be scaled up for closed circuits in retirement community and long haul trucking circuits. 

Eventually they could be highway-ready, and finally, there could be self-driving passenger vehicles suitable for all roads.

What's happening: Some of the first milestones of autonomy have already been achieved, like geo-fenced, low-speed autonomous driving, automatic emergency braking, and repetitive tasks in mining. 

In the heavy equipment industries, such as mining and construction, it's common to use AVs for repetitive tasks like moving material from one destination to another.

Autonomous drones used to monitor construction projects, handle land surveys, and watch over resource management have crossover applications in business services such as delivery, agriculture, and security.

May mobility is currently looking at use cases and testing in controlled environments, like a shuttle service in Rhode Island.

Flashback: Public expectations for passenger-ready, fully autonomous vehicles don't have a strong historical analogy. 

While smartphones come with fewer safety concerns, they followed a common pattern. The first smartphones did enough advanced tasks well enough to gain popularity. Then, as updates to software and hardware improved performance, gradually became "the hub of everything [people] do online." 

The bottom line: The best way to ensure that autonomous vehicle technology is passenger-ready before it goes into production may be to continue improving incrementally.

Bibhrajit Halder is the CEO of SafeAI, an AV technology startup focused on industrial AVs. He is also a member of GLG, a platform connecting businesses with industry experts.

Autonomy milestones in other sectors could build trust in AVs

Successful deployment in specific scenarios, like mining, could be scaled up.

Illustration of a car made out of a bunch of different cars

Instead of trying to design fully automated vehicles from scratch, the way forward might be to adapt limited technologies that have already proven themselves in specific settings, and gradually add capability.Why it matters: Adapting new technology from successful use-cases where safety is paramount, like in mining, could assuage public concerns about AVs, and accelerate AV development and production.

AVs have to interpret sensor data, determine their next moves and follow through on them, all of which requires exceedingly complex AI. To meet those demands in real time, computing for AVs will have to happen onboard the vehicle.

Why it matters: The alternative to onboard computing for driving functions would be vehicles relying on unstable network bandwidth for cloud computing while cruising at highway speeds. A specialized AI chip market has emerged to create platforms that can perform these complex computations almost instantaneously, while using as little power as possible.

Where it stands: More than 70 companies have entered the AI chip market in the last few years, and another 100 chip startups have been announced. The next generation of chips could have an array of uses, beyond advancing AV computing power. 

Nvidia is dominant in the GPU space. It offers a computing platform with a plethora of chip offerings for deep learning that take advantage of GPUs' massive parallel computation abilities. Next-generation technology, however, will introduce more efficient chip architectures designed expressly for deep neural network computation rather than graphics processing.

Google has developed a circuit called a Tensor Processing Unit (TPU) that has run common neural networks 15–30 times faster than a comparable GPU, and used far less power to do so.

Graphcore, a startup, develops accelerators and software framework together, which arguably combine to make the fastest and most flexible platform for AI applications. Graphcore has developed an Intelligence Processing Unit (IPU), which offers promise for AV use.

What to watch: Roughly 12–15 companies are pulling ahead with their next-generation AI chips, but solidifying a lead will require producing and scaling them as well, which could favor established players or motivate smaller companies to merge with larger ones, as Nervana Systems and Intel did.

How a new generation of AI chips could speed up AV computing

They could help to end reliance on cloud computing.

Illustration of a car with loading symbols instead of wheels 

Artificial intelligence

Automation and AI

AVs have to interpret sensor data, determine their next moves and follow through on them, all of which requires exceedingly complex AI. To meet those demands in real time, computing for AVs will have to happen onboard the vehicle.Why it matters: The alternative to onboard computing for driving functions would be vehicles relying on unstable network bandwidth for cloud computing while cruising at highway speeds. A specialized AI chip market has emerged to create platforms that can perform these complex computations almost instantaneously, while using as little power as possible.

The future of AVs is a whirl of hype and uncertainty, but AV technology has been used in mining and construction for decades — an often overlooked point of reference.

Why it matters: Developers of passenger AVs should be working more closely with the heavy equipment industry, which has learned key lessons about developing functional, fully autonomous vehicles, and how to prepare people for the technology.

Background: Caterpillar's Autonomous Mining Trucks have moved over 1 billion tons of material and traveled nearly 22 million miles since they were first demonstrated in 1985, with zero safety incidents.

The trucks improved efficiency and productivity by 14%, mostly in highly repetitive work.

They also removed human drivers from potentially dangerous assignments, which improved overall mining safety as well.  

Komatsu, similarly, has used autonomous mining trucks for nearly as long, and recently surpassed 2 billion tons of material hauled autonomously. 

Between the lines: Ironically, Caterpillar's success with autonomous mining trucks owes less to technology than to people. In general, a mine is a highly structured place, where people follow strict safety protocols. Anyone who might interact with the trucks was educated on how to operate safely alongside them.

Passenger AV companies cannot feasibly train all people who might come into contact with AVs. But they can provide educational materials on their websites, and work with city and state governments to offer free public education programs and to integrate AV safety into driver's ed.

Meanwhile, the technology being developed for passenger AVs, from navigation algorithms to sensors and simulation tools, is more advanced than what is currently used in mining and construction. Some of that technology could be used to develop new industrial AVs for use in warehouses, airports or shipping yards, places with more complex variables and obstacles.

The bottom line: Mining and construction have used AVs for decades without safety issues, and though passenger AV technology has much to offer industrial AVs, there are lessons that can be shared in both directions. 

What passenger and industrial AVs can learn from each other

AVs have been used in mining and construction for decades — an overlooked point of reference.

An autonomous truck staged for loading. Photo: Caterpillar

Industry

The future of AVs is a whirl of hype and uncertainty, but AV technology has been used in mining and construction for decades — an often overlooked point of reference.Why it matters: Developers of passenger AVs should be working more closely with the heavy equipment industry, which has learned key lessons about developing functional, fully autonomous vehicles, and how to prepare people for the technology.

Sections

Axios Local

Axios gets you smarter, faster with news & information that matters.

About

Subscribe

Bibhrajit Halder

Stories

Autonomy milestones in other sectors could build trust in AVs

Too much proprietary tech could hurt the AV market

How a new generation of AI chips could speed up AV computing

What passenger and industrial AVs can learn from each other