INSTITUTIONAL PARTICIPANTS

Ms. Hyoin Kim

Ph.D. candidates

Seoul National University

MS. Hyoin Kim is a Ph. D. student in the School of Mechanical and Aerospace Engineering, Seoul National University, South Korea. Her research interests include the path planning and control of unmanned aerial vehicles such as multi-copters. More specifically, her recent focus has been learning-based motion planner.

In May, 2015, she was presented the Budding Women Researcher Award by the Institute of Control, Robotics and Systems (ICROS) and Women in Science, Engineering and Technology (WISET), Korea. Her paper presented the sampling-based motion planning teachnique for mobile robot. Also, In May, 2017, she was one of the best student paper finalist in ICRA 2017. The paper proposed the learning-based motion planning framework which finds efficient motion to be used in online.

Learning-based motion planning technique: extension of parametric dynamic movement primitives (PDMPs)

Cooperation between multiple aerial vehicles can be a solution to overcome the issues of limited payload. Multiple aerial vehicles can carry a heavy or big object safely by sharing the load or relieving the risk of putting considerable inertia on a single vehicle. In reality, deploying multiple aerial manipulators in a cooperative manner is not easy. Assuming that a tracking controller can robustly follow the carefully-designed desired trajectories, a remaining important issue for cooperative aerial manipulation is to design good trajectory for each aerial manipulator. Our research, interests is a motion planning approach for complex robot task including cooperative transportation using aerial robots.

Our research utilizes the motion representation algorithm, parametric dynamic movement primitives (PDMPs), which are powerful algorithms that encode multiple demonstrations and generalize them. By learning the relationship between known environments and corresponding optimal motions, our algorithm produces adequate trajectories for aerial manipulators in new environments real-time. Since we handle the time-consuming process in offline works and simply employ the offline works in online, our algorithm is able to generate online motion in real-time. Moreover, our research includes the method to manage the demonstration set in PDMPs when some demonstrations are poor in terms of safety. By eliminating unsafe demonstrations with the parameters based on the safety criterion, and replacing them with new safe ones, we incorporate safety in the PDMPs framework. Now, our research goal is to extend the proposed motion planning framework to be used in more complex task which should employ the segmentation process. All the algorithms above have been validated in real-time experiments.