How to buy an Automated Vehicle in 2020
Starting in June 2018, the Smart Mobility Living Lab set about specifying and procuring a fleet of vehicles to be part of our facility. This fleet would include automated vehicles expected to contribute to the development and testing of:
1. ‘Enablers’ to support CAVs, such as communications infrastructure, environment mapping and road-side sensors.
2. Automated driving systems, including hardware testing and performance bench-marking.
3. Services based on CAVs, such as ride-sharing and deliveries.
To find a vehicle that would meet these needs, we spent months scouring the marketplace and talking to potential suppliers. This allowed us to learn a great deal about both automated vehicles in general, and their state of development. Now in August 2020 with our own vehicles delivered and in use, we have reviewed our experiences, and summarised that accumulated knowledge along with our perception of the present state of the industry. We hope this guide can help other individuals and organisations facing similar challenges.
1. What are the parts of an automated vehicle /self-driving car?
From a simplified viewpoint, the main puzzle pieces are:
The base vehicle: usually an established production car, but some firms are developing new purpose-built platforms.
The sensor suite and computation: typically some combination of cameras, lidar, radar, ultrasonic, and GPS sensors linked up to a computer that is sufficiently robust and powerful to consistently process the sensor outputs.
The Automated Driving System: the brain of the vehicle that sits in the computer and processes sensor outputs to determine:
o Where the vehicle is (localisation)
o What potential obstacles are in the world around it (perception)
o How the vehicle will reach its destination (navigation)
o Its choice of route through the environment immediately ahead (path planning)
o The actions to be taken to follow that path, such as controlling steering, acceleration and braking (control)
The Drive-by-Wire system: a combination of actuators and software that takes the outputs of the ADS controls and allows it to act on the vehicle. If an established production car has been used as the base vehicle, existing systems and hardware (e.g. ADAS such as automated braking and lane keeping) can be ‘hacked’, or brand new actuators may be added to sit alongside them.
Communications equipment: hardware that will allow the vehicle to communicate with the world around it (e.g. V2X) and any higher-level system managing or supporting its navigation, journeys and operations (e.g. Fleet Management Systems). More advanced and higher bandwidth communications provide the opportunity to augment vehicle performance by sharing some sensing, processing and intelligence with other vehicles and fixed roadside infrastructure.
Sourcing some of these pieces in isolation can be straightforward, but mature Automated Driving Systems are not available to purchase off the shelf. Just as challenging is the work to fit these pieces together in a way that results in a safe and capable vehicle and system.
2. Who are the players?
2.1 American Stars
There are a few firms that get the most attention in the Automated Vehicle sphere, and for good reason. Waymo (formerly the Google Self Driving Project) is the pioneer in the USA, if not globally. They have the most miles of automated on-road driving recorded, the largest fleet of vehicles and one of the most advanced ADS. Close behind in are the likes of Cruise Automation and Aptiv.
On the face of it, these firms have all the puzzle pieces: they’re working towards SAE Level 4 and have been demonstrating mature ADS on well-built vehicle platforms. However, that doesn’t mean any of them will build and sell you a car.
The vast sums of money in this industry mean that big firms’ PR setups are as advanced as their software development. What is demonstrated in a carefully crafted press release may be more impressive than is the reality. Despite their extensive city-wide trials in Phoenix, Waymo anecdotally still struggle with some key aspects of urban driving and it may be some time before they venture into more complex environments like London. Increasingly, firms have been announcing that public ready operations are further off than originally indicated.
Most big ADS developers are aiming to use their systems as urban mobility services, operating large fleets rather than selling vehicles to individuals. There are many firms competing for this goal and there will only be so much room in the market when it reaches maturity. Hence it is highly unlikely that any firm will want to sell a vehicle to an individual or external organisation as it would just be a distraction from the ultimate prize.
2.2 Traditional OEMs
With the big ADS firms capturing the limelight, it’s easy to think that traditional OEMs aren’t in the competition. However, most have Automated Vehicle Development programmes, either largely developed internally (Nissan, JLR), or enhanced through acquisitions (GM-Cruise, Ford-Argo.AI) and through partnerships (Hyundai-Aurora). Like the ADS developers, these efforts are mostly building towards service provision with SAE Level 4 systems. It is conceivable that the history and stability of traditional OEMs makes them more reluctant to oversell on their capabilities, and hence they may catch up on and even surpass the likes of Waymo.
Many companies have already been selling vehicles with SAE Level 2 capabilities to consumers. Tesla has become infamous for some of the exploits of its drivers on highways. Audi was attempting to bring the first SAE Level 3 vehicle to market with their new A8; alas they recently shelved these plans due to the complicated issues of liability in this grey area of automation.
OEMs offer the most straightforward means of acquiring a reliable vehicle platform with some level of automation. However, they share some of the limitations of the American Stars:
Vehicles with SAE Level 4 systems are still some way off the market and are likely to be channelled towards fleets operated by the OEMs or their partners.
SAE Level 2 (and 3) systems on production cars are very limited in where they can operate; they may help you parking or traveling at speed and in traffic jams on motorways, but they won’t support the bulk of urban driving.
With a mind to future commerciality, some OEMs have taken a more system-centric view. To provide a vehicle with enough sensors and computing power to operate in isolation is a very expensive prospect. Building and maintaining a fleet would cost an order of magnitude more. However, in urban areas the density of vehicles and surrounding infrastructure creates an opportunity to share the burden. Hence firms like Nissan are investing in inter-vehicle communications and off-vehicle processing to support an SAE Level 4 use-case that is more financially competitive and may offer a faster route to market.
2.3 New OEMs
What many people likely picture when thinking of automated vehicles is a purpose-built Pod or Shuttle like those often tested in cities for the past half-decade. Their futuristic aesthetics and lack of a steering wheel help to capture the imagination. The 12+ person Shuttles of French firms Navya and Easymile have been most widely used worldwide, and Aurrigo and Westfield have been operating in the UK for the past few years, such as in the Gateway Project.
The novelty of these vehicles has supported their fast-track to real-world use; their low speed and non-threatening appearance makes it more palatable to allow them into spaces usually off limits to vehicles, such as pedestrianised high streets. This in turn means they can operate without the complexity of interactions with other vehicles, significantly reducing the level of risk. However, this is also their main limitation.
While 2-4 person pods may remain acceptable, shuttles are probably too large for pedestrianised areas and will ultimately need to operate on-road with other traffic. To date this has only been feasible on private sites or with significant exceptions made by local authorities to allow them onto the road in relatively controlled circumstances. As yet, none of these vehicles appear to have cleared the many regulatory hurdles in the UK, so they cannot be used on our public road for the foreseeable future.
ADS capability typically leaves much to be desired; most trials only involve running the vehicles on “rails” on a pre-defined route, without scope for more intelligent navigation.
2.4 Tier 1s
Tradeshows are a great place to see new ideas in the Automated Vehicle Sector. In recent years the Consumer Electronics Show (CES) has had some very eye-catching stands and demonstrations, including those of Tier 1 suppliers such as Bosch, Denso and the Aisin Group. Their vehicle concepts presented a very exciting picture of what consumers may experience in the new world of automated vehicles.
However, anything seen at a trade show needs to be taken with a pinch of salt, especially where it is a concept shown off by a firm that typical only produces small portions of production vehicles. Some firms like ZF are developing whole vehicles that could soon be seen on the road. However, many others are focusing on showcasing the wide range of technology and services they are developing to sell to other firms. Hence it is less likely that they will produce a vehicle platform or ADS that will ever be available to buy.
2.5 Local ADS Developers
Around the world, there are countless smaller firms developing their own ADS. To have a chance of competing with the likes of Waymo and Aptiv, they may take more innovative approaches and may be more open to collaboration. There are a few such firms in the UK:
The most prominent are Oxbotica and FiveAI, who have been trialling on-road for a few years. The former has been active in Oxford, the latter around Millbrook, and both have more recently been active around London.
Wayve started out on the streets of Cambridge with their bright blue Renault Twizy and a vision-based machine-learning approach to their ADS.
Fusion Processing has been using their CAVstar system in numerous platforms, ranging from the Westfield POD in the Gateway Project to the CAVForth Bus.
AVL, experienced in developing ADAS for OEMs, are leveraging these skills to produce their own ADS which they are integrating into vehicles themselves.
We have found some of these firms more approachable and open to considering supplying and implementing their ADS in novel circumstances. However, such collaboration still has its limitations.
Some firms may take a “black-box” approach where the focus is on delivering a particular service or capability, at the expense of future flexibility and visibility of data-flow in the ADS. For some cases this may be sufficient, but for general research purposes may be too restrictive.
As with the American Stars, there are limitations on the present level of automated driving capability and range of environments where vehicles are validated to operate.
2.6 Open Source ADS
In contrast to most ADS being proprietary and closely guarded, a small number are taking the opposite approach. These Open Source ADS appear to have made the gamble that opening their code to community development could lead to faster progress, wider adoption and ultimately more commercial opportunities.
An example is the Autoware platform, backed by companies such as Tier IV and Apex.AI. Another is Baidu’s Apollo platform.
These ecosystems are typically free to access and build upon, reducing the barriers to entry for newcomers. The surrounding community offers a wider pool for problem solving and development, and a shared technical background represents an easier route to future standardisation and business ventures.
However, to date these open platforms have mostly been used for limited and isolated trials by smaller organisations rather than larger scale deployments, and hence there is less support available to help a new user get started with the system. These platforms are also a long way from having a complete set of capabilities validated for on-road urban driving. A vehicle with a suitable architecture will also still need to be sourced.
2.7 Conversion Services
Building a safe and reliable autonomous ready vehicle is a daunting task, so firms such as StreetDrone and AutonomouStuff have set themselves up to take on the challenge for you. They typically start with an off-the-shelf production vehicle and integrate the additional hardware needed for automated driving. In some cases, they will also install and provide initial training with ADS such as Autoware. This ensures that the customer starts out with a safe and capable base vehicle. However, each vehicle is effectively a brand-new prototype and economies of scale are limited. You also need to be prepared to continue developing and managing the ADS yourself, as typically there little long-term handholding beyond hardware service agreements.
3. SMLL’s Solution
It soon became apparent to us that highly-automated vehicles were a long way from being available for purchase for organisations like SMLL, let alone consumers. Coming out of this lengthy assessment of the market, we re-framed our priorities to sourcing:
An Automated Driving System with sufficient data accessibility, so that we didn’t restrict our scope for future research
A suitable development pathway rather than a complete offering
A vehicle and Automated Driving Systems from the same supplier, as SMLL and its partners lacked the experience to carry out the integration
A versatile electric production-vehicle platform with a robust and mature drive-by-wire system
A solution that maximised learning and development opportunities
It is for these reasons that we chose StreetDrone as our first vehicle supplier. They were well placed for integrating our various puzzle pieces which included:
Nissan ENV200 Combi – this base vehicle was the only fully-electric MPVs available in the UK
StreetDrone’s Xenos Drive-by-wire system – first established on the Renault Twizy
Autoware.ai – StreetDrone integrated and developed Autoware’s R&D-focused ADS for on-road use in our test-bed
Validation and Safety Case – StreetDrone and TRL collaborated to ensure that Autoware’s implementation was safe and reliable
A range of flexible hardware supportive of communicating with other vehicles, our data-centres, and our extensive roadside infrastructure
SMLL now owns two such vehicles, and in a trial in October and November 2019 we made the first UK public road deployment of an Open Source Automated Driving System. This achievement was only a stepping-stone to future projects. We look forward to our vehicles providing our customers with testing and data gathering platforms on which to develop and benchmark their innovative hardware, software and services. This could be self or manually driven, with exposure to all the complexities and opportunities our city has to offer.
So it may be that you don’t need an automated vehicle of your own, and our vehicles could serve your purposes. If this is of interest, or you want to learn more about our experiences, please don’t hesitate to get in touch.
By Adam Gristock – SMLL Technical Consultant