|
Finite Memory Loosely Coupled Integration on GPS/INS Integrated Navigation System |
|
Duck Hyun Suh, Dong Kyu Lee, Choon Ki Ahn(Korea University, Korea) |
| Global position system (GPS)/inertial navigation system (INS) integrated system provides a more precise position for many applications, such as vehicles, aircrafts, and mobile phones. However, this system often causes the error accumulation of inertial measurement unit (IMU). This problem may result in providing an inaccurate position. To solve such a problem, we propose a novel finite memory loosely coupled integration (FMLCI) method on the GPS/INS integrated navigation system. |
|
|
24/7 Elderly Guard Robot: Emergency Detecting, Reacting and Reporting |
|
Deok-won Lee(Gwangju Institute of Science and Technology, Korea), Ahmed Elsharkawy(GIST, Korea), Kooksung Jun, Yun-dong Lee, SeungJun Kim, Mun Sang Kim(Gwangju Institute of Science and Technology, Korea) |
| As the number of elderly persons increases, greater attention must be given to how they or their caregivers deal with emergency situations. This paper describes an automated tracking, fall detection, and emergency recovery system for elderly persons, and shows that efficient a Socially Assistive Robot (SAR) can resolve emergency situations and abnormal behaviors for at-risk populations. Our assistant robot uses position data provided by Ultra-WideBand (UWB) wireless network and motion sensor information to detect potentially dangerous situations for elderly persons. In this context, a deep neural network–based double-check method has been developed to detect and confirm fall situation. |
|
|
Finite Memory based Secure Estimation of an Unmanned Ground Vehicle for Reliable Cyber-Attack Estimation |
|
Hyun Ho Kang, Sang Hyeon Oh, Choon Ki Ahn(Korea University, Korea) |
| In this paper, we propose a finite memory based secure estimation algorithm (FMSE) to accurately estimate cyber-attack of an unmanned ground vehicle (UGV). The modeling of UGV is reformulated by defining an augmented measurement vector on the horizon. The proposed algorithm is designed on the reformulated system with feedforward structure that shows the finite memory structure under the unbiased condition. Due to the finite memory structure, FMSE shows the accurate performance and robustness under inaccurate noise information or corrupted measurements. Frobenius norm is introduced as the cost function to solve the gain of the proposed algorithm. |
|
|
Deep Reinforcement Learning Control of Spider Robot using Proximal Policy Optimization Algorithm |
|
Abdul Manan Khan(Hanbat National University, Korea), Hoan Quang Le, Buhyun Shin, Bongjo Ryu, Youngshik Kim(Hanbat National University, Daejeon, Korea) |
| In this paper, we have presented simulation results for our spider robot. We have developed reinforcement learning control using Proximal Policy Optimization (PPO) algorithm. Simulation confirms that after 1000 iterations, mean standard deviation converges which shows that PPO learn to walk using spider robot. In future, we would implement the presented algorithm on experimental system and analyze its performance. |
|
|
A Development and Verification of Steam Turbine Speed Controller in a 1,000MW Rated Coal Fired Thermal Power Plant |
|
Woohyun Ra, Inkyu Choi, Mansu Shin(KEPRI, Korea), Youngbok Han, Sanae Kang(Korea Midland Power Co, Ltd., Korea) |
| Prior to the initial steam admission to 1,000MW steam turbine in a thermal power plant after its construction, the turbine control system was verified by using a process model developed based on the heat balance and actual process operating data of same type thermal power plant. The total system where the process model was integrated into the
controller hardware with input and output modules is generally called verification system. Unlike a widely used simulator composed of only personal computer and software in the laboratory, this verification system have a great difference which
can be connected with actual speed control system and field devices by use of actual signal lines. |
|
|
Deep Reinforcement Learning of 2-DOFs Legged Hexapod Robot |
|
Hoan Quang Le, Abdul Manan Khan, Wanghun Lee, Youngshik Kim, Hocheol Lee, Buhyun Shin(Hanbat National University, Korea) |
| Autonomous robots used for rescue or exploration needs to work in unknown environment. Such robots should select appropriate actions corresponding to their environments. One of the main challenges lies in the difficulty of controlling the multi-legs of the robots with coordination to a complex dynamic environment. In this research, we develop 2-DOF legged hexapod robot getting better actions in unknown environment with using Proximal Policy Optimization - based reinforcement learning (PPO-RL). At first, the kinematics analysis of the hexapod design is presented and the results of all formulation and learning control are applied to the prototype, and then verified. This paper studies learning |
|