The NTNU CIR Lab : Moving Robots from Self-Moving to Self-Learning

A+ A- go back

The wave of artificial intelligence (AI) technologies has brought new elements to robotics. The National Taiwan Normal University Computational Intelligence and Robotics (NTNU CIR) Lab is using advanced AI design and application frameworks to move robots from merely automatic, to new horizons in self-learning.

Members of the CIR lab were interviewed by the MOST Center for Global Affairs and Science Engagement, in which they shared how the research team is helping join academia and industry to create interdisciplinary collaboration.

Interdisciplinary collaboration; moving robots toward self-learning

Three faculty in the Department of Electrical Engineering, NTNU, serve as co-CIR Lab supervisors: IEEE Fellow Prof. Wei-Yen Wang; IET Fellow Prof. Chen-Chien Hsu; and Prof. Hsin-Han Chiang. The Lab works hard to design and develop learning from human demonstration (LfD) robotic systems– that is, helping robots evolve toward self-learning. The Lab’s areas of research include AI systems, neural networks, fuzzy systems, evolutionary algorithms, digital control systems, mobile robot navigation technologies, and LfD robotic systems.

Prof. Hsu explained that, in the traditional factory, most robots precisely and repetitively perform specific tasks via predefined programs or teach panels. But as soon as the work environment changes, the engineers have to re-program or manually guide the robot to make adjustments. Not only does this take a lot of labor, but it also increases building costs. But the trend toward low-volume automation systems means that, if we can develop more intuitive designs that give robots autonomous cognition abilities, the robots would then be able to adapt to the changes more spontaneously.

Prof. Wang added that the CIR Lab was established in early 2007, with a research focus on robotic self-learning. In 2017, AI applications began receiving attention around the world; and “deep learning” has now become the watchword of the day.

How does one teach a robot to learn? Prof. Wang emphasized that smart machines require extending into and integrating other domains if the development is to proceed apace. As a result, not only has the Lab team actively sought out collaboration opportunities on campus with other programs in electro-optical and vehicle engineering, they have also done multidisciplinary research with the College of Education and College of Arts, NTNU, helping robotic learning process as similar as possible to human intelligence.

The world's smallest AI chip – to make electronics even more intelligent

In the past three years, there have been 19 research projects and highlight results completed by the CIR Lab’s three supervising professors, including LfD robotic systems, autonomous tracked robots, and more. Their research projects have received more than NT$33 million in funding. Prof. Wang said that the CIR Lab’s specialty is teamwork; many of the group’s innovations have been the result of getting the whole group’s heads together and letting the sparks of creativity fly. When describing the team’s collaboration, Prof. Hsu quoted former National Chiao Tung University President Chang Mau-Chung’s famous words: “Act Together, We Go Far!”

One of the Lab's accomplishments particularly worth noting is that, in 2018, they worked together with a chip producer to create the “AI Mipy” chip, measuring only 7x7 mm, making it the world’s smallest AI chip. In 2019, the team announced an “AI development board for end-user products”, which can be successfully integrated into a wide variety of electronics to directly grant artificial intelligence to electronic products.

Prof. Wang stated, the AI algorithms that exist today all have to undergo programming with a desktop or other large computer to reach effectiveness. If that process can be reduced to a chip, then it will no longer require laptops or high-performance computers; instead, the process will be able to operate independently, thus reducing costs and lowering energy usage. Thus, future equipment will need only a single chip to complete AI tasks, which will represent a major step forward. The currently available development board uses a camera as its input, with functions focused on image recognition. In the future, it will have applications in AI monitoring to help make factories smarter, help make up for labor gaps in long-term care facilities, and more.

The trend towards human & robot collaboration; strengthening talent development

Taiwan is entering an aged society with low birth rates. We are also living in an era in which the COVID-19 pandemic has broken up global supply chains, an era defined by labor shortages and opportunities for contactless business. In such an era, having AI robots with cognitive abilities will mean being able to better adapt to different production environments, and helping humanity accomplish tasks that are difficult or even impossible for humans to accomplish.

What will the next stage in the development of smart machinery hold? Prof. Hsu puts it this way: Humans and machines working side by side is the trend of the future. Collaborative robots will be able to work alongside humans, taking on more agile and flexible tasks; as well as meeting the need for low-volume and high-variety production. For that reason, guaranteeing the safety of robots on worksites will become a critically important issue. Next, the CIR Lab will continue to deepen gesture recognition, natural language processing, adaptive grasp strategies, and other research topics. The Lab’s goals in the next stage are to continue to optimize the interaction between humans and robots, enhance productivity, and use sensor fusion technologies to build a safe working environment for human-robot collaboration to avoid workplace accidents.    

Long-term, stable development for the robotics industry will also depend on talent development. Prof. Wang pointed out the amount of emphasis the CIR Lab places on integrating theory with practice. This means that, by participating in research project competitions, graduate students can develop solid practical abilities, and build greater links with industry. He hopes that the lab’s research outcome can be also integrated into the industry's forward-looking research development, so the competitive technologies could create academic, industrial, and entrepreneurial value.

Go Back