How Smart Cloud Infrastructure Helps a Self-Driving Company Pave the Way for Seamless Partnerships
The name of the game when it comes to deploying autonomous vehicles is matching the supply of vehicles with customer demand. For Argo, that demand comes from establishing partnerships with leaders in both ride hailing and delivery services that have already built up strong networks of customers. But, there needs to be a matchmaker to connect the supply with the demand, and for that, we’ve developed ArgoWatch, a cloud-based system that monitors and manages all aspects of Argo’s autonomous vehicle fleet and makes it easy for us to plug in with partners.
ArgoWatch’s interface is designed to seamlessly integrate with go-to-market partners’ digital systems, allowing us to get up and running quickly with new services, like the self-driving ride-hail operation launching with Lyft, or upcoming retail delivery services.
Communication in the cloud
ArgoWatch provides a single point of integration for our partners, like Lyft, so they don’t have to communicate directly with our individual vehicles. ArgoWatch enables us to pre-position our vehicles in high-demand areas, like outside concerts or sporting events, ensuring that vehicles are ready to pick up potential Lyft customers and get them where they need to go quickly.
Each Argo vehicle is equipped with a high-speed LTE connection that enables them to instantly send relevant data to and from the cloud. As well as managing routing and route planning, we monitor resources like fuel levels and battery charges; we keep a careful eye on the state of the autonomy system (for example, how much onboard data storage is left) and we can support each vehicle during its journey. This data pipeline also enables a specific vehicle–or the entire fleet–to receive information about unexpected road closures, construction work, or even traffic backups spotted by on-board sensors, and adjust accordingly.
The phantom fleet
To ensure our operation is always improving, we deploy our phantom fleet – a simulated fleet that drives around virtual representations of the real-world cities where our vehicles test — continuously learning and gathering data about the dynamics of our cloud and (virtual) fleets.
As the phantom fleets – consisting of anywhere from one to 1 million simulated vehicles – make their way around the virtual streets, they receive ride-hail requests from virtual riders just as our test vehicles do in the real world. This methodology is helpful to provide oversight from an operational perspective, with the virtual testing team simulating ride-hail requests, cancelling rides, and changing rider plans en route, all under varying supply and demand conditions.
We take scenarios observed out on the public roads across our six test cities–everything from wrong-way drivers to out-of-nowhere construction zones–and build them into virtual testing. We can even run an event that occurred in one city on the virtual streets of another city to see how the system would respond.
Pressure testing the system
While the phantom fleet is a major tool in our testing arsenal, it’s not the only way that we use simulation at Argo. Along with extensive public road testing, simulations help ensure our self-driving system (SDS) can handle as many scenarios out on the road as possible, including the all-important edge cases—rare situations that autonomous systems need to be ready to handle. Our goal is to operate our fleet safely in every condition and under every scenario that the real world may throw at us.
We run all these tests to identify the limits of our technology within our geonet, or the network of roads that constitute our launch area for our commercial partnerships, like with Lyft. We need to ensure that the system can efficiently allocate rides at the busiest times of the day, while ensuring there are no software bugs or poorly allocated vehicle assignments along the way. And we simulate at scale, to ensure that the system will always ensure optimal routes and fleet positioning, even at times when the streets are flooded with ride-hail requests.
We cannot afford to leave a passenger stranded, or an inefficient vehicle assignment that doesn’t allocate the closest vehicle to a requesting ride-hail passenger.
Plug-and-play for partnerships
The flexibility of the phantom fleet and other virtual testing helps enable Argo to simulate various types of partnerships—we can operate cars with multiple partners at the same time, and even use a phantom fleet vehicle in ride-hail mode for a period of time, and then switch it to goods delivery.
Finally, as we prepare to launch commercial services powered by ArgoWatch, virtual simulation enables us to speed up integration with new partner systems, by testing and confirming that their software is compatible with our systems before we use it to run cars on the road. If it works with our phantom fleet, we can confidently test it in a fleet of real cars.
With an eye on scaling, we’re currently running the phantom fleet in tandem with real-world testing in Miami, Austin, Pittsburgh, Detroit, Palo Alto and Washington DC, with preparations underway to launch fleets in Hamburg and Munich in Germany. We also developed a custom-designed app for Argo employees in each of these cities to test our self-driving system with Ford vehicles in a ride-hailing service.
When combined, our fleet management system and the ride-hail test app—effectively the backstage and front of house interfaces of ArgoWatch—open the door to wide-scale commercialization of the Argo self-driving system. Together, it unlocks the ability to optimize fleet position, assign or alter routes, and monitor critical resources such as fuel, battery charge status, and other vehicle functionalities.
These are linchpins to a future of trusted, reliable, and enjoyable self-driving, which Argo aims to scale for a diverse set of partners in the years to come.