Cisco and DAI-Labor Have Transformed Berlin's Historic Thoroughfare into a Test Track for Autonomous Vehicles

CISCO

The self-driving car is the holy grail of automotive technology. Many efforts to create an autonomous vehicle have focused on building intelligence into the automobile itself, thus transforming it into a moving nerve center. But this approach is based on old-fashioned thinking that seeks to replace a human operator with a machine.


What if we made the car a component of a wider intelligent system? Why stop at equipping it with an independent computer brain and sensors that scan and react to the outside world when we can transform our cities and roadways into smart networks that interact with the vehicle?


This idea may sound revolutionary but is, in fact, evolutionary. You are reading this on a device connected to the internet, which is a distributed system of networked computers that communicate, coordinate, and pass messages to one another to provide services like email, chat, gaming, and streaming.

Distributed Computing and the Internet of Things

The Internet of Things—or IoT—is extending this distributed computing model into the real world by combining cloud technology, remote sensors, and edge network devices to collect and process real-time data generated by physical events, actions, and objects.


In its simplest form, IoT powers smart speakers and thermostats in homes, self-checkout machines at supermarkets, heart rate monitors on smartwatches, and tracking devices affixed to merchandise and personal belongings. But these applications only scratch the surface, and the industrial IoT is revolutionizing everything from agriculture to transportation.
 

Governments around the world are partnering with public and private research facilities to build better communities. These smart cities are using sensors and other IoT technologies to collect data and manage municipal assets, resources, services, and infrastructure.
 

I run DAI-Labor, the Distributed Artificial Intelligence Laboratory at the Technical University of Berlin. Our facility focuses on research and development in intelligent services and systems that can help meet the challenges of today and tomorrow.
 

Our experts work in diverse fields that include distributed computing, machine learning, network security, and interactive services. We strive to produce research results that are touchable and realizable, and our work has resulted in a wide range of software tools and frameworks, models, and methodologies that we've deployed outside the university. 

User-Centered Research and Development

DAI-Labor's R&D activities are user-centered. We focus on user requirements at the design phase and deploy our applications in realistic environments to thoroughly evaluate whether they meet end-user expectations and needs. Our presentations and demonstrations target academics, technologists, and the general public. We strive to make our research accessible to the widest possible audience.

Impactful AI research should reach the widest possible audience.

At any given time, we are developing dozens of projects that will have an immediate and a future impact on the well-being of millions of people. Our current initiatives include the Bobbi Chatbot, a virtual assistant that provides authoritative information to citizens of Berlin, freeing up live operators to assist people with issues that require human intervention and problem-solving skills.


Drones4Life is optimizing the delivery of medical specimens and blood products to hospitals. It uses autonomous airborne vehicles (drones) instead of cars and trucks to lower costs and to speed delivery times by creating direct routes and eliminating bottlenecks due to traffic.


RouteCharge is building a network of battery exchange stations at 150 km intervals to increase the range of electric trucks by providing dual-use facilities that charge and swap vehicle power packs.


These ongoing projects are in various stages of development, but they have moved beyond the laboratory and are presently deployed in the real world.

Our Autonomous Vehicle Platform

One of the most exciting initiatives is DIGINET-PS, our autonomous vehicle platform that is currently being tested right here in Berlin.


DIGINET-PS is a digitally connected protocol route that takes a new approach to vehicle autonomy. It decentralizes and distributes computer power between cars, digitized road objects, and the cloud.


On-board vehicle sensors, including cameras, lidar, and radar, interact with roadside units to record, process, and share data, leading to the early detection and avoidance of risks. Outdoor devices include traffic light sensors that detect oncoming vehicles and calculate the duration of signals, road condition sensors that monitor surface conditions for dangers like black ice and potholes, and environmental sensors that monitor weather and pollution levels. 

Building a Secure Urban Test Field

We have deployed these technologies in an urban test field that comprises 3.7 kilometers of road, six lanes of traffic, two multi-lane roundabouts, 15 traffic light systems, and various parking situations. Our trial environment stretches from Ernst-Reuter-Platz to the Brandenburg Gate along Straße des 17. Juni in the heart of Berlin. It was chosen to reflect the complexities of driving in the German capital.


The Brandenburg Gate is a cultural monument and a tourist attraction, and so this is a heavily trafficked route. It is also used by diplomatic and government convoys—for example, during state visits by foreign dignitaries—and is closed to vehicle traffic during celebrations like the annual New Year's Eve party, and during political rallies and protests.


At the moment, we are using six different types of roadside sensors along the test route. These devices have to talk to one another, communicate with moving vehicles, and transmit information to our data center for further processing. This facility employs some 4,000 cores, and 10,000 GPUs to convert the data collected by cars and our edge network into actionable information.

  

We partnered with Cisco to build a secure IoT platform that integrates these sensors and provides a secure network that cannot be hijacked by external entities.
 

Cisco roadside units collect data from sensors along the route, and Cisco Kinetic for Cities provides middleware that manages our IoT devices and edge network, and powers our distributed system. We use Cisco UCS to manage and secure our data center, and rely on the company as a go-between that leverages its global reach to connect us to various technology partners.


For an ambitious project like DIGINET-PS, you need the right teammate. We’ve found that in Cisco. Cisco’s value truly is the ecosystem that they bring—both in technology and in other partners. 

Our Four Goals

We have four goals with this project. Our first goal is to build the software component that integrates the vehicle, the roadside edge network, and the cloud. Secondly, we are providing an API that will enable entrepreneurs to build and test solutions that work with the DIGINET-PS platform. Our third goal is to partner with OEMs, like German carmakers, to test their vehicles' autonomous driving capabilities.


The fourth and final goal is equally important. As part of the Technical University of Berlin, we are providing colleagues working on other projects with infrastructure that will help them advance research into machine learning, cybersecurity, and IoT integration. 

Berlin and Beyond with Cisco Kinetic

Cisco Kinetic technology is powering smart city projects all over the planet. Berlin has joined world capitals like Paris and London and smaller cities like Raleigh, North Carolina, in bringing the digital transformation to public services and urban infrastructure.


DAI-Labor is expanding the DIGINET-PS project. Right now, we have two autonomous cars on the road in the test area. We'll soon be adding an autonomous bus with displays that will allow passengers to interact with the vehicle using a chatbot. We are also extending the test track by ten kilometers.


Our eventual goal is to establish a national standard for autonomous vehicles and smart thoroughfares. We are already working on that with the German government, but first, we must expand DIGINET-PS to other cities, and that means laying down more roadside infrastructure. And that means working with Cisco to replicate our success in Berlin.


The distributed approach of DIGINET-PS addresses one of the biggest limitations of self-driving cars that use onboard systems alone. Driving conditions can change in an instant. Intelligent roads equipped with cloud-connected sensors can interpret and ferry data to approaching vehicles to provide a failsafe when autonomous cars don't have time to react or encounter situations that can fool their built-in programming.


For example, a car's cameras may fail to detect a truck that is braking or changing lanes because they cannot distinguish its low-contrast paint job or the reflective surface of the container it is hauling from the sky. However, roadside sensors can send a signal warning about the movements of this object that onboard technology cannot see.


In another instance, a self-driving car programmed for a specific route may not register a concrete barrier blocking access to an off-ramp that is under construction. In this case, a roadside beacon can serve advanced notice causing a vehicle to change its itinerary, or it can send telemetry data that will trigger an emergency maneuver to prevent an imminent collision.

Autonomous cars need smarter roads #CiscoKineticforCities #TUBerlin #DAI-Labor


It is my wish that DIGINET-PS and our RouteCharge network of vehicle battery replenishing stations eventually expand to cover all of Germany's roadways, making our country a leader in green and safe automotive technology.


We also hope to bring our solutions to the international community, thus improving transportation options for people around the world. The road ahead is wide open, and I am confident that the engineers and researchers at DAI-Labor and Cisco will continue to work together to map the future.