BrandPosts are written and edited by members of our sponsor community. BrandPosts create an opportunity for an individual sponsor to provide insight and commentary from their point-of-view directly to our audience. The editorial team does not participate in the writing or editing of BrandPosts.
By Debra Slapak, Florian Baumann
I remember the tiniest details of my first plane flight, from San Antonio, TX, to Yuma, Arizona. Back then, we bought our tickets from a travel agency and carried traveler’s checks rather than cash or credit cards. I passed through Phoenix, where a manual security check of my purse identified pepper spray—which I learned I would be leaving behind. Only a couple of flights per day connected Phoenix to Yuma, and I got my connection information from an agent holding a printout of schedules and gate numbers. But even back then, those schedules and gate numbers were computer-generated in an age when most things weren’t.
The transportation industry has a long history of harnessing the latest technologies to deliver better mobility experiences and services. Today is no exception: The industry is undergoing a sweeping digital transformation that enables new ways to move about — from connected and autonomous vehicles to pay-per-use and mobility-as-a-service models.
Enabling these new mobility models can be challenging for even the most advanced organizations in the industry, often due to the shortage of skills to envision and implement new data-driven ways of solving age-old problems. But the rewards are tremendous:
Manufacturers can build safer vehicles that reduce the chances of deadly accidents.
Mass transit operators can keep subways, trains and buses up and running on schedule.
Airlines can ensure greater security and offer personalized services to soothe weary travelers.
Municipalities can reduce congestion on their streets and highways to keep the flow of traffic moving at more predictable rates — and reduce the pollution that comes with gridlocked roadways.
All these use cases are made possible by putting massive amounts of data to work in systems driven by artificial intelligence (AI). From the automotive design shop to the urban planning office, AI is the foundation of a new era of connected mobility.
AI use cases
Perhaps the most common use case we hear about is AI that makes split-second decisions in autonomous vehicles and advanced driver-assistance systems. These advances incorporate a wide range of AI-enabled technologies — such as deep learning, natural language processing, computer vision and gesture-control features — to provide the “intelligence” for vehicles to operate safely.
Many of the vehicles on roadways today are “intelligent.” They include features such as emergency braking, cross‑traffic detectors, blind‑spot monitoring and driver‑assisted steering. These vehicles are at what we would call “Level 2” in the diagram below. Meanwhile, across the industry, instances of Levels 3-4 are ramping quickly within controlled environments.
At a broader level, AI isn’t just changing our automobiles; it’s redefining how we think about mobility. For example, AI-driven applications are behind the shift from privately owned vehicles to mobility as a service (MaaS). With MaaS models, users choose the best form of transportation for the moment — from rideshares and cars used as a service to subways and buses. This isn’t a vision of the future; it’s a description of what is happening today. A report by the American Public Transportation Association describes how cities such as Vienna, Hamburg and Helsinki integrate the MaaS concept. For example, in Helsinki, residents can use a single AI-supported app on their personal devices to plan and pay for train, subway, bus, streetcar, or ferry rides.
Similar changes are taking place across the broad realm of transportation infrastructure. In many cities, smart traffic routing based on real-time information from sensors along roadways helps transportation officials reduce congestion and gridlock on streets and highways. Meanwhile, cities rely on AI algorithms to optimize bus and subway schedules for a better passenger experience.
And, of course, the industry is putting a lot of focus on AI-driven predictive maintenance. These applications leverage data from IoT sensors in connected vehicles to monitor the performance and condition of equipment and to watch for and predict component failures. This continuous intelligence allows manufacturers, owners and operators to work proactively to keep cars, trucks and public transit systems up and running at an optimal level.
Top considerations for capitalizing on AI
As diverse as these use cases are, they share common threads– considerations for successful implementation. To highlight these considerations, I turned to Dr. Florian Baumann, Chief Technology Officer for Automotive and AI at Dell Technologies. He outlines a handful of crucial recommendations for organizations that want to develop, test and continually refine their AI uses cases and algorithms.
A fully automated toolchain
“For starters, to successfully deploy complex AI-enabled workflows that require continuous integration, organizations need a fully automated toolchain,” Florian says. “This toolchain is key to streamlining and accelerating development lifecycles.” He explains that a toolchain brings essential development capabilities together into a single tool– including data collection, data ingestion, data preparation, algorithm training, algorithm test and validation, and production—and reduces time-to-market.
“In general, we see three possibilities to leverage toolchains for AI development,” Florian notes. “You can build your own toolchain, you can use open-source tools, or you can use commercial solutions. With all the available options, you can find an AI toolchain that supports your organization’s specific requirements.”
Multi-site, multi-party enablement
Continuous integration and continuous delivery in iterative processes
Florian notes that organizations also need to consider the iterative nature of AI development and the need for continuous integration and continuous delivery (CI/CD). The development process is not “one-and-done.” The task consists of a “repeated process to train, test, retrain and improve the algorithm on an ongoing basis, to ensure the trained AI model maintains its accuracy as the world–and the data fed into the AI model–evolves and changes over time. This iterative process requires integration and delivery throughout the development cycle. An architecture that encompasses all these sub-processes is essential for speeding up the development project.”
Database and metadata management
“As they face continuing data growth, organizations also need more sophisticated tools for database and metadata management, along with extremely scalable storage systems and advanced tools for moving, processing and managing massive-scale datasets,” Florian says. “These tools are essential for gaining insights from petabytes of data.”
High performance computing systems with easy management
And, of course, there are hardware and security considerations. Training deep learning algorithms with today’s vast, complex datasets is computationally intensive. AI training is essentially a big data problem that requires high-performance computing systems with the latest CPUs, optional accelerators, fast interconnects, high-performance storage. Given the importance of data, organizations need to implement robust solutions for data protection, data security and data privacy. And all of this infrastructure needs to be easy to orchestrate and manage.
“Because people skills and talent are difficult to find nowadays, data scientists must be able to concentrate on developing and modeling algorithms, instead of wrestling with the complexities of the underlying infrastructure,” Florian explains.
AI is a force that is enabling a fundamental shift to a new era of digitally-enabled mobility, with nearly limitless opportunities for enterprises and public organizations that are poised to capitalize on AI use cases.
How do they get there? As organizations move forward, Florian suggests concentrating on the fundamentals.
“Focus on scalability, from the beginning on,” he says. “Focus on easy management of software and hardware. Leverage end-to-end toolchains and fully integrated solutions to keep your data scientists and the rest of your team productive and highly motivated. And remember that data management and metadata management are critical to success.”
Call to action
For a broader and more in-depth look at the transformational power of AI in the transportation industry, see “The New Mobility Industry.” This paper includes insights into the foundational steps organizations can take to achieve success in the new mobility industry — from adopting a data-centric strategy to enabling the workforce of the future.