System Bits: June 18

Another win for aUToronto

Photo credit: University of Toronto

The University of Toronto’s student-led self-driving car team racked up its second consecutive victory last month at the annual AutoDrive Challenge in Ann Arbor, Mich.

The three-year challenge goes out to North American universities, offering a Chevrolet Bolt electric vehicle to outfit with autonomous driving technology. The goal is to have fully self-driving EVs in 2020. This year, the competition emphasized urban driving. The eight teams had to demonstrate how their vehicles could respond to traffic signs, traffic lights at intersections, and pedestrian crosswalks. The final challenge took place at MCity, a simulated town course at the University of Michigan for testing self-driving vehicles.

“I am extremely proud of our team,” said faculty advisor Tim Barfoot, a professor at U of T’s Institute for Aerospace Studies (UTIAS). “It was fantastic to see all their hard work pay off. The car performed pretty much exactly as planned. The team really came together and did a brilliant job of supporting each other.”

Photo credit: University of Toronto

The aUToronto team paired engineering students with computer science students.

“We performed perfectly on the traffic sign challenge, with a flawless parking job at the end,” says technical team lead Keenan Burnett, a U of T engineering science graduate who’s now pursuing a master’s degree at UTIAS. “During the MCity challenge, we made it the farthest out of all the teams.”

Burnett says the key to their win was their many enhancements to U of T’s self-driving car, Zeus, over the past year. These included the use of LiDAR (light detection and ranging) localization to precisely determine the car’s position, and employing deep neural networks to correctly detect pedestrians and traffic lights. To do this, they collected a data set of more than 70,000 images.

The team placed first in all but one of the nine judging categories. They received top marks in social responsibility, the mapping challenge, pedestrian challenge, MCity challenge, and more. They were runners-up in concept design.

“We are very proud of the work we’ve done so far and have our eyes on the future,” says Burnett. In preparation for the final competition next year, the team will begin improving Zeus’ object-detection capabilities and focus on developing a fully reliable and safe operating system.

U of T competed with Kettering University, Michigan State University, Michigan Tech University, North Carolina A & T State University, Texas A & M University, the University of Waterloo, and Virginia Tech.

“I think we have a good shot at year three,” adds Burnett. “We hope to test Zeus on public roads sometime in the next year.”

Caltech uses AI for smoother drone landings
Landing a drone with multiple rotors can be challenging. Due to the turbulence created by the rotors, landing and takeoff can present operational issues, especially with autonomous drones.

The California Institute of Technology’s Center for Autonomous Systems and Technologies (CAST) came up with the Neural Lander, a system with a deep neural network, combining artificial intelligence with control technology, to work out the wobbles in drone landings.

“This project has the potential to help drones fly more smoothly and safely, especially in the presence of unpredictable wind gusts, and eat up less battery power as drones can land more quickly,” says Soon-Jo Chung, Bren Professor of Aerospace in the Division of Engineering and Applied Science (EAS) and research scientist at the Jet Propulsion Laboratory, which Caltech manages for NASA. The project is a collaboration between Chung and Caltech AI experts Anima Anandkumar, Bren Professor of Computing and Mathematical Sciences, and Yisong Yue, assistant professor of computing and mathematical sciences.

Deep neural networks (DNNs) are AI systems that are inspired by biological systems like the brain. The “deep” part of the name refers to the fact that data inputs are churned through multiple layers, each of which processes incoming information in a different way to tease out increasingly complex details. DNNs are capable of automatic learning, which makes them ideally suited for repetitive tasks.

To make sure that the drone flies smoothly under the guidance of the DNN, the team employed a technique known as spectral normalization, which smooths out the neural net’s outputs so that it doesn’t make wildly varying predictions as inputs or conditions shift. Improvements in landing were measured by examining deviation from an idealized trajectory in 3D space. Three types of tests were conducted: a straight vertical landing; a descending arc landing; and flight in which the drone skims across a broken surface—such as over the edge of a table—where the effect of turbulence from the ground would vary sharply.

The new system decreases vertical error by 100%, allowing for controlled landings, and reduces lateral drift by up to 90%. In their experiments, the new system achieves actual landing rather than getting stuck about 10 to 15 centimeters above the ground, as unmodified conventional flight controllers often do. Further, during the skimming test, the Neural Lander produced a much a smoother transition as the drone transitioned from skimming across the table to flying in the free space beyond the edge.

“With less error, the Neural Lander is capable of a speedier, smoother landing, and of gliding smoothly over the ground surface,” Yue says. The new system was tested at CAST’s three-story-tall aerodrome, which can simulate a nearly limitless variety of outdoor wind conditions.

“This interdisciplinary effort brings experts from machine learning and control systems. We have barely started to explore the rich connections between the two areas,” Anandkumar says.

Besides its obvious commercial applications—Chung and his colleagues have filed a patent on the new system—the new system could prove crucial to projects currently under development at CAST, including an autonomous medical transport that could land in difficult-to-reach locations (such as a gridlocked traffic). “The importance of being able to land swiftly and smoothly when transporting an injured individual cannot be overstated,” says Morteza Gharib, Hans W. Liepmann Professor of Aeronautics and Bioinspired Engineering; director of CAST; and one of the lead researchers of the air ambulance project.

This research was funded by CAST and Raytheon Company.

CMU’s work on lithium-ion batteries
Carnegie Mellon University’s Venkat Viswanathan and graduate students Shashank Sripad and Dilip Krishnamurthy are working to improve the safety of lithium-ion batteries, which are common in devices such as smartphones, hearing aids, and digital cameras.

Viswanathan, an assistant professor in mechanical engineering, researches how designing materials can create novel energy conversion and improve storage capabilities.

Lithium-ion batteries “are basically the most energy-dense batteries you can find,” Sripad said. “For many portable electronics, they are the only battery that you can use.”

Unfortunately, they also have been known to unexpectedly short out, causing fires and other safety issues.

These shortages are caused by dendrites, miniscule branches of ions in the interior of the battery. When a battery charges, lithium ions travel between the positive and negative end, also known as the cathode and anode. Sometimes the ions don’t travel into the anode, instead depositing on its surface. They stack over time, creating dendrites. If the dendrites reach the cathode, the battery shorts out and a fire can start.

Viswanathan and his team hope to improve battery safety by preventing dendrite growth at its source – the anode. They are focusing specifically on dendrite growth in lower temperatures. When the battery’s environment is cold, “the energetic barriers become more important,” Krishnamurthy said. “Meaning it will take not be as easy to push the lithium ions into the anode.”

In order to prevent barriers, the “battery spends its own energy heating up,” Sripad said. Needless to say, this uses up energy and decreases the battery’s effectiveness.

For this research, Viswanathan received a 2019 Young Investigator Award from the Office of Naval Research. The Young Investigator Program supports young academic scientists and engineers in researching topics that could benefit the goals of the Navy and Marine Corps. As the recipient of the award, Viswanathan will work closely with other Naval Research Laboratory researchers, to make the work relevant for naval applications.

Viswanathan and his team said they hope that the collaboration will result in a practical way to mitigate dendrite growth. If they succeed, the improved battery safety could have a tangible impact in both the Navy and the public sector.

“Think of submarines, or even cruise ships,” Sripad said. “All of them are exposed to low temperatures. This will definitely improve the reliability and safety of most of their energy storage systems.”

For Viswanathan and his team, the coming years will likely hold challenges and triumphs. And lots and lots of batteries.