WPI Symposium Challenges Robot Industry, Academics to Come Together

Robotics displays and presentations were part of a WPI symposium on robotics. Image: Joe Dolen

October 30, 2019      

WORCESTER, Mass. – Members of the robotics ecosystem – educators, industry leaders, and students – recently gathered at a symposium hosted at Worcester Polytechnic Institute. During “Launching to a Robotic Future”, speakers gave updates on robotics research, panels discussed bridging the gap between researchers and business, and students showed off their latest work in the robotics field.

The event was held last week, days before WPI announced it had secured a five-year, $3 million grant from the National Science Foundation to focus on research and training related to the adoption of robotics assistants in the workplace. The symposium was also a celebration of sorts – honoring the 10+ year history of the robotics program at WPI (the program started in 2007, but awarded its first four bachelor’s degrees in 2009).

The program at WPI is “an incredible model for higher education, and for the nation to not just talk about working across silos and boundaries, but actually obliterating them and creating something new in its place,” said WPI President Laurie Leshin in her opening remarks. “ You’ve clearly had your finger on the pulse of what the market wanted from us too, because there’s no program at WPI that’s grown larger or faster in the last decade.”

Bridging the gap

The symposium featured two panel sessions that discussed how researchers and businesses in the robotics space could work better together, to help bridge the gap in the so-called “valley of death.”

“Should research be application driven or discovery driven? I would say it has to be both,” said Chris Jones, the CTO at iRobot, a speaker on the “Robotics Research Directions and Applications” panel. “If you look at it holistically around how does innovation happen, there does need to be a discovery phase where it’s much more open-ended and there is a fundamental trying some new things to see what can be shaken free. But ultimately for those innovations to have an impact, it needs to become application driven, where it’s grounded out in a real product – something that has a sustainable impact in terms of its interaction with the world. That tends to be a course where an industry is a more central figure.”

WPI symposium panel Scott Dresser Amazon

Scott Dresser, director of software, systems and solutions at Amazon Robotics, moderated a panel on Robotics research directions and applications at the WPI robotics symposium. Image: Joe Dolen

Jones explained that there’s a “fuzzy middle” between the research and business communities, which requires that both groups come to realize that it’s not a one-time technology transfer opportunity.

“In terms of how the sides work together, you need both application and discovery,” said Jones. “Industry needs to have the patience and foresight to work with academia to try to understand where this is going and how to harvest and translate that into an application. But also from the research side, to realize that this takes time and to make an effort as things progress, to be more involved in trying to articulate how this could impact an application area and become engaged in more specific obligations to help make that transfer. Navigating the fuzzy middle ground is a hard part, though.”

Mathew DeDonato, a manager in the vehicle hardware group at the Toyota Research Institute, as well as a WPI alumnus, spoke about creating a balance between the worlds of academia and industry. “The research really enables new applications and opportunities that can go back to develop new research,” said DeDonato. “For example, Toyota Research Institute wouldn’t exist if there wasn’t an end goal to make money at the end of it for Toyota. They’re putting a lot of money into our company so we can be on the cutting edge of the future technology. Keeping that healthy balance between the two really grounds both sides. On the other side, is that researchers will never really let something out of the lab, it will never truly be finished if there isn’t someone pushing them to pull it out of the lab for the real world.”

Academic keynotes

The WPI symposium also featured three academic keynote sessions from professors at other universities, giving their take on the latest technologies and research.

Dr. Robert Howe, the Abbott and James Lawrence Professor of Engineering at the Harvard Paulson School of Engineering and Applied Sciences. Howe spoke about the development of robot hands that attempted to mimic the human hands.

“Even back in the 1980s we were able to make a robot hand that copied the kinematics and structure of the human hand,” said Howe. “However, controlling it in a useful way has proven very difficult. You can see wonderful human-like motion, but it’s teleoperated – every joint motion is directly transferred from the human.”

It turns out that attempting to recreate a human hand for a robot hand was proving to be difficult, Howe said. “So we decided about 15 years ago to try a new direction,” he said. “Let’s try to keep things simple and we asked how far can you get if you just use one motor and really minimize the sensors that forces you to use passive mechanics.”

After creating a prototype and figuring out the mechanics of a hand, Howe’s team was able to develop robot hands that could grasp objects without having to duplicate the intricacies of a human hand. This led to further development and companies like RightHand Robotics, which was co-founded by a team of researchers that included two members from the Harvard Biorobotics lab.

Howe said research within the sensors space have allowed his research to investigate additional opportunities and applications for robot grasping. “This is a golden age for robot sensing,” said Howe. “It’s amazing now what these off-the-shelf sensors allow you to do. We can take advantage of that as well. The pressure sensor is going to be key, and we’re working hard in that direction. I can’t claim victory at this point, but I think the availability of those sensors is going to definitely tick up what robot hands can do in the real world.”

History lessons

Two other academic sessions focused on the history of some of the robotics technology that we take for granted in 2019. Dr. Wolfram Burgard, a professor from the University of Frieburg, Germany, who is currently the vice president of automated driving technology at the Toyota Research Institute, spoke about his early work in probabilistic and machine learning in the autonomous driving space. Dr. Shigeki Sugano, professor and dean at the School of Creative Science and Engineering at Waseda University, Japan, talked about the origins of robotics all the way back to Greek mythology, and how the development of robots also mirrored creative and artistic views of robots in science-fiction literature and movies. Sugano highlighted a lot of the university’s work on creating humanoid robots that can assist humans in assistive care situations.

 

Spot Boston Dynamics WPI Symposium

Spot, a robot from Boston Dynamics, made an appearance at WPI’s Robotics Research Symposium. Image: Joe Dolen

The final keynote session was from Dr. Al Rizzi, the chief scientist at Boston Dynamics, and included an appearance by the company’s new Spot Mini robot that can walk, climb over difficult terrain, and perform other cool tricks.

Rizzi gave attendees a background of the work from Boston Dynamics and its early research for the military that included on figuring ways to transport soldiers’ gear that could maneuver over difficult terrain. From the development of the “big dog” robots to the more maneuverable Spot robots, to the Atlas humanoid robot, Rizzi discussed how the company progressed from being a research-focused organization to become more of a product-focused organization while continuing to do research.

WPI symposium Al Rizzi

Dr. Al Rizzi, chief scientist at Boston Dynamics, discussed robotics developments at the company. Image: Joe Dolen

There’s still a lot of interplay between what we’re doing in terms of pure research and things that we’re doing as demonstrations on Spot, which we consider a product at this point,” said Rizzi.

As an example, Rizzi showed the audience a YouTube video of an Atlas robot doing a tumbling routine, explaining that the team took about a month to author those behaviors, compared to a similar video from almost two years earlier in which Atlas did a back flip. In that example, it took the team about three months to figure it out. “We’re pretty optimistic because the ideas we’re using here to do this with Atlas actually play a big role in helping us author behaviors for all the machines we’re working on more effectively,” said Rizzi. The Atlas robot, primiarly an internal research project, “is an opportunity to work on pretty far out ideas and make sure we can realize them on hardware,” he added.

The ideas can go back and forth from the research side to the commercial side, Rizzi explained. An example of where Spot could jump up and down was carried over to the Atlas team, and they were able to push it further to create more exciting moves.