In the Loop

European universities swept the SpaceX Hyperloop finals.

Imagine sitting in a capsule that’s traveling at 1,100 km (700 miles) per hour through a steel tube in a partial vacuum, a few meters off the ground, like a person-size version of the pneumatic mail tubes that used to carry letters around buildings. Jump on in London and be in Edinburgh in less than 30 minutes. Depart Paris at 09:00, and you’ll be in Barcelona by 10:00.

The TUM pod in the test track, ready for launch (Image: TUM)

This concept, conceived by Elon Musk in 2012, is called Hyperloop. The idea is to combine low air pressure with a pod that floats above its rails on air bearings in order to reduce air resistance and friction, the main factors that make current surface transport inefficient and slow.

In his original whitepaper, Musk proposes that Hyperloop’s optimum utility is for linking pairs of cities no more than about 1,500 km (900 miles) apart; farther than that, and it’s faster and cheaper to fly.

Commercial efforts to develop Hyperloop technology are proceeding by Musk’s SpaceX company and others. Virgin Hyperloop One has a full-size test track in the Nevada desert, where the Virgin team has tested a prototype pod at up to 240 mph. Hyperloop Transportation Technologies and TransPod are not far behind. Here in Europe, Dutch company Hardt Hyperloop aims to open a test center by 2021.

To encourage development of the technology, Musk held the fourth annual Hyperloop contest for teams of engineering students at the SpaceX 1.2-km test track in Hawthorne, California, this summer. Hundreds of entries from universities around the world were narrowed down to four finalists, all from European universities: TUM Hyperloop, from the Technical University of Munich; Swissloop, from ETH Zurich; the EPFL team, from École Polytechnique Fédérale de Lausanne; and Delft Hyperloop, from Delft University of Technology.


The team from the Technical University of Munich won the competition for the second consecutive year. The TUM Hyperloop pod reached a maximum speed of 463 km/h (288 mph), just under the previous world record.

The TUM team chats with Musk. (Image: TUM)

Under the rules of the competition, the pod that reaches the highest speed without crashing wins. This year, SpaceX increased the demands on the teams by adding requirements for communications systems on the pods (in previous years, SpaceX provided the comms), and a requirement to stop within 100 feet of the far end of the tube. This could take the form of a single main run to that point or a “slow crawl” after the pod’s main run was completed.

TUM’s pod was based on the winning design from last year, keeping the eight brushless-DC (BLDC) motors but using smaller motors with a maximum combined output of 320 kW, for twice the power of the 2018 design. Like the earlier design, the pod measures 1.7 m long and 50 cm wide, but at 70 kg, it weighs 8 kg less than last year’s pod. This means a favorable increase in the power-to-weight ratio.

“The main change [compared with last year’s pod] was in the power supply concept. The battery compartment is more compact and has less weight,” said team member Sophia Ramirez. The pod uses lithium-ion–polymer battery cells, heated to improve their performance.

The design uses hundreds of parts from sponsor Infineon, including 36 MOSFETs and 24 Hall-effect switches for current commutation in each of the eight motors, plus a further 112 MOSFETs in the battery main switches.

The sensors used in the pod’s systems include analog thin-film pressure sensors deployed in the brakes. Navigation was achieved with a combination of Hall-effect sensors to measure the rotation of the wheels and diffuse sensors used to detect stripes of tape placed every 100 feet inside the test track tube for this purpose. Counting the number of stripes passed verified the position of the pod.

“For the communications system, microcontrollers communicate with each other via CAN [a Controller Area Network bus],” said team member Tim Klose. “The telemetry board of the pod translates the CAN messages into UDP [User Datagram Protocol] packages, which are transmitted to a ground station — a laptop outside the tube — via a network provided by SpaceX.” The biggest challenge the team faced in building the pod was time management. “Since a new pod has to be designed every year, which is necessary according to SpaceX rules, time is always short,” Klose said. “Also, only a limited number of test runs on SpaceX’s test track are possible, and in these test runs the optimal configuration has to be found.”


Delft Hyperloop took fourth place in this year’s contest with a maximum speed of 196 km/h. Chief Engineer Elja Ebbens explained the unique challenges of the Hyperloop environment to EE Times Europe.

A member of the Delft Hyperloop team makes final preparations for its pod’s race-day run. (Image: Delft Hyperloop)

“The high-speed regime of more than 600 km/h on a track with up to 1-mm bumps poses an immensely high load case on the entire vehicle and especially on the wheels, at more than 40,000 rpm,” he said. “All the parts on the vehicle have to be able to withstand these loads and vibrations expected during a run. At these speeds, it is no longer possible to copy from existing vehicles. We had to invent and design everything ourselves — literally, reinventing the wheel!”

Ebbens also pointed out that in a vacuum, components begin to arc. Dissipating heat from electronic components in vacuum conditions is also particularly difficult, especially given that motors and batteries are being pushed close to their temperature limits.
The Delft pod also runs eight BLDC motors: four at the front, four at the back. This configuration is one of the biggest changes from last year’s Delft design, which featured one large motor and a single large battery pack, said Ebbens.

“The smaller motors perform substantially better, but we needed eight to meet our torque requirements,” he said. “With smaller motors, it was also easier to test one to destruction in order to provide invaluable data which allowed us to push the motors to their physical limits.”

The Delft Hyperloop team celebrates with Musk. (Image: Delft Hyperloop)

Four battery packs power two motor subsystems each. These operate in vacuum conditions, unlike last year’s design, which had the battery in a heavy sealed box to keep it at atmospheric pressure. Each battery pack comprises four custom-made modules in a 2S2P configuration (two cells in series, then a pair of those connected in parallel).
“Within these modules, lithium-polymer [LiPo] cells are laser welded together through copper plates, allowing for a low-resistance connection. Using modules, we can easier replace faulty modules and limit the voltage we work on,” Ebbens said. “For each battery pack, we use a custom EMUS battery management system to monitor each individual cell for their voltage and 25% of the cells’ temperatures, along with the pack discharge current, and to automatically balance the cells. In total, we use almost a thousand LiPo cells on our vehicle to power our motors.”

The suspension system is combined with the drivetrain to ensure all eight vertical guidance wheels can deliver torque. The navigation and control subsystem includes five CPUs, including an Arduino Uno (required by SpaceX for data readings), a Raspberry Pi for communications, and three STM32 Nucleo boards, each with a custom PCB acting as interface board and shield for the Nucleos.

Each Nucleo board has its own function: sense, think, or act. The “sense” board handles incoming sensor data from the pressure sensors in the brakes and measures the ambient pressure and temperature. “Think” runs the navigational algorithm based on motor controller readings and diffuse sensor readings that count the tapes passed in the tube; this data is used to estimate the correct brake point before the pod hits the end of the test track. “Act” activates all eight motor controllers with a target pulse width modulation (PWM) signal and controls the brake calipers.

“Using three CPUs for these tasks meant we could increase the debuggability of the software and decrease the cable lengths on the vehicle to reduce weight at the same time,” Ebbens said.

The Raspberry Pi handling communications was connected to a 900-MHz radio, which sends data to the ground station outside the tube via a radiating cable positioned overhead along the length of the tube. Keeping a steady connection with this cable in vacuum conditions (which don’t allow for much heat dissipation) during the run proved to be too much for Delft’s communications system on race day: The connection was lost, causing the emergency brakes to be applied before the pod could reach its top speed.

Overall, the Delft team was proud to have reached the final round of competition, given the strength of competition from the other entries as well as the challenging time and testing constraints, Ebbens said. “The entire vehicle has to be made within one year, which was a real challenge by itself. Even if you did have time to test before the competition, you are not able to test your vehicle in the competition conditions, in the 1.2-km vacuum tube, before that one final run on competition day. Each vehicle can only be run once in these conditions, and in this one run it has to work perfectly to win.”


Musk has said that next year’s competition will use a 10-km test track — more than five times longer than the current SpaceX test track — with a curve. The inclusion of a curve will make the competition a more realistic test of Hyperloop technology, but this will of course make it much harder for the student teams. Traveling around curves introduces lateral forces, which increase with the square of the pod’s speed; at hundreds of kilometers per hour, this is not to be taken lightly.

Europe’s finest technical universities are betting they will prove once again that their engineering students are up to the challenge. ■