|
Development of Safety Diagnostic Robots with Artificial Intelligence for Industrial Plants |
|
Jun-Hyeon Choi, YeChan An, Sung-Hyeon Joo(Sungkyunkwan University, Korea), Tae-Yong Kuc(Sung Kyun Kwan University, Korea) |
| In this paper, a robot for safety diagnosis is presented to solve the problem of safety diagnosis at industrial plant sites. Safety diagnosis is essential for industrial plants because accidents at industrial plants cause huge economic losses as well as casualties. However, safety diagnosis is difficult due to various dangers such as suffocation and gas explosions in industrial factories. Therefore, we have developed a robot that performs safety diagnoses instead of humans. |
|
|
Disturbance Observer-based Robust Path Tracking Control for Tracked Vehicle |
|
KwanKyun Byeon(Chung-Ang University, Korea), Junghwan Gil(Univ. chungang, Korea), Sesun You, Hojin Lee, Wonhee KIM(Chung-Ang University, Korea) |
| In this paper, we propose path tracking control for tracked vehicle. Tracked vehicle model is modified to linear model using lumped disturbance which includes nonlinear terms and external disturbance terms. The model is separated two subsystems, upper system and lower system. State references are given by upper controller via backstepping. Lumped disturbances are estimated through disturbance observer (DOB). Lower controller is designed to define error dynamics using state references. Furthermore, estimation disturbances are compensated through DOB in lower controller. The stability of the proposed method is proved through Lyapunov theory. |
|
|
QR-SCAN: Traversable region Scan for Quadruped Robot Exploration using Lightweight Precomputed Trajectory |
|
Eungchang Mason Lee, Donguk Seo, Jinwoo Jeon, Hyun Myung(KAIST, Korea) |
| Quadruped robots are facing the contact constraints of the legs, namely traversability. In this paper, to explore unknown environments with quadruped robots safely, we propose a novel local exploration planner that utilizes the precomputed trajectories to check traversability and collision at the same time. By strictly checking the collision and traversability in two steps, the admissible path is guaranteed in the receding-horizon manner. In addition, the trajectory considers the kinodynamics and is lightweight thanks to its precomputation. The performance of the proposed method is verified by exploring 3D simulation environments in comparison with one of the state-of-the-art methods. |
|
|
Deep Reinforcement Learning based Planing for Urban Self-driving with Demonstration and Depth Completion |
|
Chuyao Wang, Nabil Aouf(City, University of London, United Kingdom) |
| Research shows major interests in urban self- driving in recent years, both perception and motion planning considered to be significant topics. Current techniques of decision making for driving policy are modular and hand designed, which is expensive and inefficient. With the development of machine learning, learning-based approaches have become a mainstream research direction. However, the performance in urban driving scenarios is far from satisfaction due to the brittle convergence property of deep reinforcement learning and debased observation. To solve these problems, this paper proposed a |
|
|
Adaptive Sliding Mode Control for Trajectory Tracking of the Tracked Vehicle System |
|
Hojin Lee, KwanKyun Byeon(Chung-Ang University, Korea), Junghwan Gil(Univ. chungang, Korea), Wonhee KIM(Chung-Ang University, Korea) |
| The dynamics and kinematics model of tracked vehicle are complicated due to nonholonomic constraints, skid-steering motion, etc.
Moreover, unknown terrain condition, external disturbance and parameter uncertainty may cause undesirable vehicle motion.
To overcome aforementioned challenges, adaptive sliding mode control (ASMC) strategy is used without requiring highly accurate vehicle dynamics model.
System dynamics is separated to body frame scheme for upper controller and velocity dynamics for lower controller. |
|
|
Learning Control Applications for Autonomous Driving in Extreme Maneuver Scenarios |
|
Son Tong(Siemens Digital Industries Software, Belgium) |
| This work presents the applications of iterative learning control for autonomous vehicle in extreme maneuver scenarios. By exploiting data from previous executions, the proposed learning algorithm that uses a simple kinematic model can generate optimal feedforward control signals to improve vehicle control performance against nonlinearity, model uncertainties and disturbance. The design developments are validated using a co-simulation structure of Siemens Simcenter Amesim and Prescan software. We validate and analyze the results with two driving use cases: racing car and drift parking. |
|