How data and analytics is helping to make the Hyperloop dream a transport reality

Profile picture for user gflood By Gary Flood December 15, 2020
Spark and Databricks functionality is at the heart of the recent successful passenger test for Hyperloop

An image of passengers sitting in Hyperloop
(Image sourced via Hyperloop )

Although many remain sceptical, Elon Musk's demand for a ‘hyperloop'- magnetic levitation-powered trains carrying passengers at high speeds - is quietly powering ahead, COVID notwithstanding. In November, Virgin Hyperloop One carried out its first passenger test and longest test trip (15 seconds) in the Nevada desert, zoning down an airless tunnel at 107 mph.

Though the ultimate aim is approximately 7 times faster than that, it's definite progress for the company which claims to be launching the first new mode of mass transportation in over 100 years. The test showcased a number of its technologies, from electric propulsion and electromagnetic levitation under near-vacuum conditions, and it says it's now working with governments, partners, and investors around the world to make the idea of connecting cities like metro stops - with zero direct emissions - a reality in "years, not decades".

‘A system that will deliver the travel experience of the future'

Intrigued by news of the test, we reached out to Jerome Wei, Senior Director of Machine Intelligence & Analytics at the company, to see what IT is going into the project. Turns out, quite a lot, especially around what he described as seeking to apply advances in algorithms and computing to realise next-generation transportation capability. He said: 

We're building a system that will give back time and deliver the travel experience of the future, because if we only invest in the same technologies we've had for more than a century, tomorrow will look like today, only much worse. It's been over a century since the Wright Brothers first showed us human flight was possible, so it's time for a new era in transportation capable of carrying us forward for the next 100 years.

So as a new technology, we are starting from the ground up with a highly digital strategy, incorporating data as a core element of system value. This data will enable a wide range of intelligent capabilities, including high reliability through predictive maintenance and high efficiency and performance through demand forecasting. Leveraging data, our fully automated system can actually react to changes in demand.

Wei went on to explain that Hyperloop is composed of a fleet of relatively small, autonomous, high speed, and high efficiency electric vehicles operating on a dedicated guideway. These system characteristics enable operational advantages in terms of fine grained management of capacity and service level, but to maximise such potential advantages, his team is developing something called the Virgin Hyperloop Command & Control System, which is made up of a number of interconnected automated planning, traffic management, and autonomous vehicle operations.

Why? Because the aim here is to enable a seamless passenger experience, allowing purchase of tickets through a mobile based app and being directed and guided to time and place of vehicle boarding within a hyperloop station (or in Virgin Hyperloop One language, a ‘portal'). Wei explained:

Behind the scenes, our automated planning system automatically generates a highly performant schedule of trips, directs the appropriate autonomous vehicle to execute that trip, and guides it from departure to arrival in a safe, coordinated manner. We have developed a detailed, physics based simulation of a hyperloop system including vehicle dynamics, sensors, communications, and guideway infrastructure. With this virtual hyperloop, we are then able to simulate and analyse the performance of our system in full scale operations and characterise its performance.

Sounds great, but how? The answer: analytics. Wei, who has also worked in areas like satellite tech for Boeing and autonomous vehicles for Northrop Grumman, said analytics has been of "critical importance" since the beginning of the Virgin Hyperloop story. He added:

Developing a highly complex system that is very different from existing technology on a highly accelerated time frame is a challenge. Transportation network level simulation, grounded in physics and real world movement patterns and dynamics, has enabled us to virtually operate and assess the performance of our design and quickly iterate and improve. With our Analytics Pipeline, we are able to transform detailed data captures to high level insight about passenger experience, safety, efficiency, and cost.

Analytics at Virgin Hyperloop enable meaningful communication between engineering and business organisations to understand system and business requirements, assumptions, and critical design trade-offs. We are able to assess, understand, and tailor our product to customers' needs, as well as support collaboration with regulators and other stakeholders.

Specifically, the project is starting to base its engineering decisions on an intelligent analytics framework based on a diverse set of data configurations such as route information, travel demand and population information to iterate on Hyperloop models faster. The plan is to also leverage public and private transportation data to optimise design and route schedules, enhance performance and improve safety and reduce data processing time.

Hard, quantitative evidence 

Wei's department is partly using open source to do this in the form of the Apache Spark unified analytics engine for big data processing, which he says was the best match in terms of processing performance and data flow needs, as well as what's essentially its commercial successor, the Databricks data engineering and collaborative data science platform. Wei notes that the latter's integrations and "out of the box usability" greatly accelerated time to first analytic product delivery, and that it continues to enable the team to stay focused on our core mission of delivering design insight. He also praises the tool for what he said were "very opportune: new feature releases that addressed project processing needs in a timely way", as well as a feature he finds useful for logging, analysing, sharing and reproducing Machine Learning or AI experiments.

As a result, he stated, his group is able to support Virgin Hyperloop One project teams in modeling, simulation, and analysis of prospective hyperloops across the world and provide detailed analyses in "weeks" instead of months. Next up: he wants to rapidly assess projects and generate what he describes as "hard, quantitative evidence and rationale" around hyperloop capability.

The company has a key focus on achieving safety certification by 2025, and our work will play an integral role in providing key system evidence for this safety and certification process.

Summing up, for Wei, as an exciting new technology company trying to solve big, world problems, it is expected that not only the technical feasibility (‘Can it work?') but the business and operational performance (‘Can it generate value?') will be questioned. But in his view, he added:

It is the responsibility of the innovator to convince the skeptics, and Virgin Hyperloop has been able to apply technology to accelerate the project delivery process for large infrastructure.

Now, with our recent passenger testing, we've proven that not only can the technology work, but that it's actually safe for humans.