The Computing and Control Engineering Laboratory (CoCEL) of the department of Convergence IT Engineering was founded in Sept. 2014 with the mission to develop futuristic AI-powered instrumentation and control systems for a safe and convenient life for those in need. From innovative fundamental ideas and theories, we realize world-changing instrumentation and control technologies through multidisciplinary research in computing and control engineering including machine learning, dynamics, embedded systems, control theory, signal processing, information theory, communication, and networks.
To achieve these missions, CoCEL has the following two vision statements: “Paradigm shifter of instrumentation and control engineering”, and “Lab of the students, by the students, for the students.”
Though instrumentation and control engineering are old and well-established, CoCEL struggles to radically deviate from existing frames and provide new technologies and new theories that are indispensable for the upcoming years. For successful research, all CoCEL members including the professor must share a strong relationship of trust beyond the perfunctory relationship between teachers and students. CoCEL actively encourages members to have core values such as generosity and humanity, high-level professionalism, positive thinking, penetrating insights, and integrity. CoCEL members should have the professional expertise to give solutions to challenging real problems that many people need. CoCEL has refrained from doing “research for research” or “research for writing papers”. CoCEL’s research work is not at the level of simulation and analysis but leads to startups by creating real prototype products.
CoCEL has three research groups working on drones, controls, and energy for specific and intensive research, with a solid grounding in computing and control engineering.
The drone research group focuses on the development of autonomous aerial vehicles. For this purpose, 3D omnidirectional depth sensors and autonomous navigation solutions have been developed and further commercialized through startups. Integrating cutting-edge technology, CoCEL is developing an ambitious autonomous nano drone. Nano drones are palm-sized. It is challenging to achieve miniaturization and autonomy simultaneously. Generally, autonomy requires heavy and oversized sensors for high performance. Such sensors are, however, not proper for nano drones due to their limited payload. The bottom line is that small and lightweight sensors are essential for autonomous nano drones. CoCEL has developed new lightweight and cost-effective 3D omnidirectional depth sensors based on laser triangulation. The developed sensors ensure a wide field of view while achieving portability and affordability. CoCEL has also studied sensor fusion to enhance synergy between two sensors and the development of new sensors.
Two companies were spun off from the drone research group. Here is a brief introduction to each company.
• Polaris3D
(http://polaris3d.co/)
Polaris3D has a business targeting out-of-stock detection systems in retail stores. Polaris3D has received investments of about 300K USD so far, with expected sales of 1M USD this year. It is supported by TIPS (Accelerator Investment-Driven Tech Incubator Program for Startup), which is an association where the Korean government invests with venture capital.
• Hybo
(https://www.hybo.co/)
Hybo is developing solid-state, compact size, super-speed, SLAM (Simultaneous Localization and Mapping) embedded, and super-wide line laser sensors for various industrial applications. In 2019, Hybo received a pre-order of 30K USD.
The control research group has developed a variety of control systems for autonomous vehicles and industrial facilities for POSCO. In particular, our research puts much weight on recent state-of-the-art reinforcement learning, rather than using classical control approaches to deal with real complicated control problems. Sensor fusion technologies with LiDARs, radars, and cameras have been critical research issues. A robot manipulator is also an important research topic for smart factories and automation. The robot manipulator is a collection of links that are connected by joints. To make robot manipulators follow the given reference trajectories precisely, these joints should be strictly controlled. However, it is typically not tractable to control them accurately because of their high nonlinearity and unmodeled uncertainties. To resolve these difficulties, we have presented the practical adaptive time delay control (ATDC) that adjusts control gains adaptively and then guarantees uniformly ultimately boundedness tracking. Recently, we are working to make the proposed ATDC work with cutting-edge reinforcement learning.
The Energy research group focuses on constructing a sophisticated model of Lithium-ion batteries to consistently monitor their internal states. Specifically, Li-ion batteries have high energy densities, high power densities, and good rechargeability with no memory effect. For the safe and efficient operation of such practical Li-ion batteries in real applications, it is essential to monitor the internal physical parameters of the batteries. Various methods have been employed for parameter identification of Li-ion battery models, but the use of approximated models, local optima trapping, and data inefficiency limits them. To overcome these issues, CoCEL used AI and machine learning approaches. To validate these approaches, CoCEL is working with Samsung SDI. Using learning-based methods, a data-efficient parameter identification scheme for Li-ion batteries is developed even with very sophisticated electrochemical models. With these elaborate battery cell models, our research will be devoted to providing diagnosis, prognosis, and control of Li-ion batteries for efficient and safe usage.
Propelling along the way we have been doing so far, CoCEL will continue to educate and research fundamental and applied technologies that are desperately needed, and nurture talented students who take responsibility for finding Korea’s next growth engine.
