Delivering Autonomy
An introduction to autonomous vehicles, where we are now, and why we’ll see widespread adoption in the delivery services first.
Introduction to Autonomy
Today, the average luxury car has around 100 million lines of software code. Compared to the 400,000 lines of code used by the original space shuttle, this is a staggering testament to how many features that cars now have. Over the past century, driving processes have been made increasingly automatic or easier to use (e.g., self-starting system, power-steering, cruise control). Many of these new features are driven by safety reasons (no pun intended), but recently, we’ve been seeing a rise in infotainment and communication (e.g., Onstar) as our cars have become more and more like our tablets, smartphones, and computers.
Looking forward, the next stage of evolution points toward autonomous vehicles (AVs). AVs have received a bit of buzz lately, as different AV startups have gone public or been acquired. But it’s also important to note that beyond the hype, there are currently no fully autonomous vehicles on the road. The levels of autonomy, as laid out by the Society of Automotive Engineers, is as follows:
- Level 0: Driver Only, driver in control at all times
- Level 1: Driver Assistance, braking and acceleration support provided
- Level 2: Partial Automation, combine L1 assistance features
- Level 3: Conditional Automation, self-driving under limited conditions and driver must take control if system requests
- Level 4: High Automation, self-driving under limited conditions (geofenced or can only drive in certain areas) but driver does not need to take control
- Level 5: Full Automation, completely self-driving and steering-wheel optional
Currently, the highest level we have achieved for passenger use is a level known as Level 2+, which includes systems such as Tesla Autopilot. This level includes 360-degree surround sensors and deep neural networks (DNN) so the system can perform more automated actions such as lane changes or highway entrances and exits. It’s a level defined by tech suppliers that reflects one of the great bottlenecks of AV evolution — because the leap from Level 2 to Level 3 removes the driver from the equation, yet many countries still require a human driver behind the wheel, current vehicle models can only expand on Level 2 capabilities through Level 2+.
Tug-of-War: Where We Are
You might be thinking — I wouldn’t even trust some people to drive, why would I trust a computer? In that, you’re not alone. A 2017 Pew Research survey of 4,135 Americans found that 53% of the people surveyed were more worried than they were enthusiastic about autonomous vehicles. When asked if they would ride in an autonomous vehicle, 56% of people said "No," with a majority saying that they either don't trust the technology to take over or they have other safety concerns. A recent MorningBrew survey also found that 75% of 1,052 survey respondents said that they’d like clarification on who would be responsible in the event of an AV-involved accident.
But really, who is responsible?
There are so many more components to a self-driving vehicle than a regular, Level 0 car, and all of these components are usually supplied by different companies. A brief overview on all the components:
- The car itself, which is either supplied by an existing original equipment manufacturer (OEM) like Mercedez-Benz, or created as a purpose vehicle like Nuro
- Mapping technology, which needs to be more precise than GPS and requires the Global Navigation Satellite System (GNSS) or crowdsourced mapping solutions like Mobileye’s Road Experience Management (REM)
- Communication technology, such as for cars to communicate with each other or infrastructure (e.g. to know when the light is red) with dedicated short range communication (DSRC) championed by NXP or cellular-based communication (C-V2X) championed by Qualcomm
- Sensors, usually a combination of cameras, radar, and light detection and ranging sensors (LIDAR) produced by companies like Velodyne or Luminar
- In-vehicle computer, powered by semiconductors and millions of lines of software code, which fuses all the sensory data together to actually drive, led by Intel and NVIDIA
As you can see, creating an AV is a joint effort from multiple different companies, and if one area fails, then it has serious implications for safety. Combine that with the mixed bag of policies between countries, and even between states in the US, and we have a regulatory mess. Only recently did Japan become the first country to allow Level 3 cars on highways. And although the US has testing hubs for AVs, mostly concentrated in Arizona and California, the uncertain public perception and policy measures have served as major roadblocks to AV evolution.
So, what now?
The hype around AVs centers around passenger vehicles, or cars that everyday consumers can drive. This is where the roadblock starts, and what brings me to a detour that we can take instead— automated delivery services.
Delivery is a huge industry by itself, and so I want to focus on food delivery. Other aspects of delivery can benefit from automation too (and probably will before passenger vehicles, for similar reasons), such as trucking or package delivery. Food delivery is unique though in how common and consumer-facing it is, which will be advantageous to speeding up the entire AV evolution timeline.
We believe the benefits of autonomy will first be realized in local commerce. Because no one rides inside delivery vehicles, the self-driving software can focus on protecting other road users, and does not need to make tradeoffs with occupant safety or comfort. An R2 vehicle could slam on the brakes when a ball rolls across the road without fear of causing whiplash, or operate more conservatively in the first place without concern for making a rider late to a meeting.
— Nuro
The food delivery industry has also skyrocketed since we all moved inside with COVID-19. A recent report from Morgan Stanley Research has found that the US food delivery TAM in 2020 is around $325 billion, and could potentially reach $450 billion by 2025. 65% of respondents to a survey run by Alphawise have also reported ordering food delivery online in the past six months.
There are a number of reasons why we will see huge AV growth in the food delivery industry, drawing primarily from a synergistic relationship. As hinted at before, these reasons include:
1. Cost Saving
Food delivery services have struggled with profitability, relying on promotions and deals to entice a largely sensitive consumer base and stand out amongst a relatively homogenous group of competitors (re: UberEats, Doordash, Seamless). AVs are a potential cost-saver, and it is estimated that every 1% of autonomous food deliveries could lead to a 1% increase in earnings. Especially with the recent legal battles fought by Uber and Lyft on whether their drivers are contractors or employees, the possibility of automating their services could be top of mind for food delivery companies.
2. Growing markets
It’s also important to note the incredible opportunity overseas, particularly in Asia. One of the most widely used apps in Southeast Asia, Grab, has just announced that it’s going public with a $39.6 billion valuation, and it has over 187 million users. Although it offers many services such as ride-hailing, online banking, etc., its food delivery service accounts for more than 50% of its revenue. This is a testament to the huge food delivery market abroad. Research firm dataSpring notes that the Asia Pacific region was the largest market for food delivery in 2019. China’s food delivery market itself grew from around $3 billion to $50 billion in the last five years, indicating an accelerating penetration rate.
One of the most recent examples of automated food delivery is Domino’s and Nuro. Domino’s was the first pizza chain to offer mobile ordering and a digital pizza tracker system (starting in 2008!). Since then, it has consistently grown revenue by over 10% YoY. Starting this week (April 13th), Domino’s partnered with Nuro, a robotic delivery company backed by Chipotle, to test delivery out in Houston. The partnership will be using Nuro’s R2 vehicle, which is the first driverless vehicle approved by the National Highway Traffic Safety Administration (NHTSA). Nuro is already delivering food, groceries, prescriptions, and other locally-sourced goods as a last-mile service. This means that it only operates within a geofenced area, and because it only transports goods, it prioritizes the safety of humans by driving at a low speed and including a protective panel at its front to absorb energy.
There are also synergies to be seen with the food and grocery delivery services.
At Refraction AI, the pandemic-induced surge in demand brought our restaurant delivery business to another level and led us to expand into grocery delivery over the summer.
— Luke Schneider, CEO of Refraction AI
3. Regulations and public perception
Because autonomous food delivery vehicles do not require human drivers, such as with Nuro’s R2, they not only provide a cost-saving mechanism but also a way for the AV industry as a whole to push the regulatory space forward. Regulations are historically behind technology — as seen with Audi’s 2017 Level 3 prototype for its A8 model, which was never released because the regulatory framework never expanded to address Level 3 capabilities. Audi CEO Rupert Standler stated then that they will offer the feature only when governments allow it and it is legally safe.
However, vehicles that do not require drivers are subjected to a different set of regulations, and can acquire exemptions. In the case of Nuro, they were exempted from including some of the required equipment of standard, passenger-carrying vehicles such as windshield wipers and side mirrors, and so were able to increase its number of sensors to improve safety. These added features can lead to faster approval for commercial deployment, which Nuro received in California a few months after being exempted. Also, because Nuro is a last-mile delivery service, it doesn’t operate on highways, which have strict regulations and could pose the greatest safety risks.
Adoption in Asia and other regions may add to competition between countries. China in particular has been accelerating its AV adoption timeline, releasing a statement in February 2020 of how they aim to create a regulatory framework for AVs by 2025. China is also home to some of the more notable AV startups, including Nio, XPeng, and Pony.ai. Coupled with their accelerating food delivery industry, China’s regulations for AVs may progress rapidly in the next few years, putting pressure on the US too.
The domino effects continue. Faster commercial deployment can lead to more public exposure, especially since food delivery is a client-facing service. As the public becomes more comfortable and familiar with autonomous vehicles, they may also become more willing to (1) purchase their own groceries, meals, etc. using delivery vehicles and (2) be more receptive to passenger AVs down the line. This last part is crucial to widespread AV adoption.
“If I had asked people what they wanted, they would have said faster horses.”
— Henry Ford
Just as automobiles must have been both exhilarating and terrifying when they were first introduced, so will be AVs. Food delivery services then, are a way of gradually introducing both the government and the public to autonomous technologies. They are also a relatively low-risk and safe method of improving AVs, which can be done much more rapidly with more widespread testing and driving opportunities. Automating food delivery is thus the next stop in the road for AV adoption, and because it is only a budding industry itself, is one ripe with change and possibility.
Current Market Landscape
There are currently a lot of delivery robots and drones being tested and deployed in small numbers, such as Amazon’s Scout vehicle or the Starship delivery robot. However, there are relatively fewer self-driving vehicles dedicated to drive on roads, which is the area we believe will have the most benefit for AV progress in general.
Delivery is also a relatively accessible niche for startups to begin focusing on, instead of trying to tackle everything from trucking to robotaxis as companies like Waymo (Google) can do. Delivery itself is also broad, and startups tend to focus on one aspect over others, such as food, grocery, or package delivery.
These startups include:
- Nuro: R2 vehicle, partnered with and delivering for Kroger, Domino’s, Walmart, first exempted self-driving vehicle by NHTSA and first granted commercial deployment by California
- Refraction AI: REV-1 food delivery robot, can operate in car or bike lanes (top speed of 15 mph), pilot testing in Michigan winter
- AutoX: Second to obtain driverless permit in California and first in China, began RoboDelivery service testing in California in 2019, RoboTaxi service (self-driving taxis) fully open to public in Shanghai and Shenzhen
- Robomart: Convenience store on wheels (store-hailing) operating invite-only in West Hollywood
- Udelv: uPod local delivery vehicle, recently announced a deal with Intel’s AV unit Mobileye to start commercial operations in 2023 and deploy 35,000+ AVs by 2028, up to 45 mph driving
We also expect to see more partnership between companies in the supply chain as well, such as with the 2019 partnership announced by Doordash and General Motors’ Cruise Automation to test last-mile food delivery services in California.
Risks
As mentioned before, automating food delivery services could lead to job losses, and subsequently public backlash. However, AV delivery also has the potential of creating many jobs, leading to a net gain. An independent economic impact study by Steer projected AV delivery to create 3.4 million new jobs per year, a number greater than all the public school teachers in the US. Nuro has seen this job growth already taking place in their deployments in Houston and Arizona, where they started hiring workers to man the retail stores, select goods, and load the goods into the vehicle.
Another risk to this space is safety, which will affect public perception and regulation. As mentioned before though, AV delivery is one of the safest forms of AV deployment, as delivery vehicles can be geofenced, at low speeds, and because they don’t have passengers, can prioritize pedestrians and other vehicles. We see this emphasis on safety in the designs of existing delivery AVs, such as Nuro, and it is continuously prioritized in the tests and improvements still being made.
One last notable risk, which can harm not only AV delivery but AVs and automobiles in general, is cybersecurity. Because an AV relies heavily on its computing system, it is also vulnerable to hacking. Although autonomous food delivery companies have been relatively quiet on this topic, cybersecurity is top of mind for many automobile OEMs, AV companies, and industry groups,and the National Highway Traffic Safety Administration (NHTSA) hosted a roundtable in 2016 with representatives from 17 OEMs, 25 government entities, and 13 industry associations. An industry-wide push driven by urgency and the rise in recent cyberattacks is expected to be seen in the future, with which autonomous delivery will follow. If autonomous delivery is tapped into first, widespread adoption of AVs in delivery services may even accelerate the necessity for cybersecurity as well.
Final Thoughts
The automobile industry is huge, estimated to be $2 trillion worldwide. If it is completely converted to AV technology, then the AV market will be huge too.
The numbers don’t tell the full story behind the urgency for this technology though. The average commuter loses about 42 hours a year to traffic, and traffic congestion is the main contributor to greenhouse gas emissions by cars, which itself is around 28% of total emissions. What is even more urgent though, is that over 1.3 million people die in car crashes every year worldwide, and 94% of those crashes are caused by human error. AVs have the potential then, to not only save time and the environment, but to save many lives.
I haven’t gotten to other aspects of AV technology such as mobility-as-a-service (MaaS) opportunities, trucking and fleet management, and over-the-air software updates, all of which are exciting and growing spaces themselves. Food delivery, and other delivery services though, are the ones that will lift the AV industry from hype to tangibility. Delivery is a growing market itself, and can provide AVs with the opportunity to operate in relatively low-risk, geofenced areas with wide exposure to everyday consumers, making it the place where we will be seeing a rise of AV technology first. AV technology in turn will boost delivery service profitability and potentially cut costs, contributing to a snowball effect of widespread adoption and deployment.
