|
The Development and Applying of Image Classification for Diamond Quality and Characteristic Analysis |
|
Sidaporn Namchan, Anuntapat Anuntachai, Nathchaya Tapasri(King Mongkut’s Institute of technology Ladkrabang, Thailand) |
| This research studies on image classification techniques to develop and apply on analysis model for diamond qualification. The purpose is to reduce time in diamond quality inspection. The researchers collect 260 images of heart and arrow cut (brilliant cut) diamond from various sources. The collected images can be separated into 2 groups which are 160 images of perfect heart cut and 100 images of imperfect heart cut. All images are processed by image classification techniques such as image rotation, image segmentation and image edge detection. These techniques distill data about heart shape a |
|
|
Reinforcement Learning-based Sim-to-Real Impedance Parameter Tuning for Robotic Assembly |
|
Yong-Geon Kim, Minwoo Na, Jae-Bok Song(Korea University, Korea) |
| Using impedance control, it is possible to control the contact force by appropriately adjusting the impedance parameters. However, The impedance parameters should be set by the user because it is difficult to accurately recognize the dynamics of the contact environment, which takes a lot of time because it should be performed whenever the assembly task changes.To this end, a reinforcement learning-based impedance parameter tuning method is proposed in this study. Since this method uses only the physics-based robotic simulation on the virtual environment, there is no risk of damaging the robots or parts and learning time can be significantly reduced. |
|
|
Design Method for Biomolecular Circuit with Plasticity |
|
Kenta Hidaka, Takashi Nakakuki(Kyushu Institute of Technology, Japan) |
| We propose a design method for a biomolecular circuit that can dynamically change its function according to repetitive input stimuli. In DNA computing, the circuit function strongly depends on the network structure and reaction rate parameters of the chemical reaction system. In other words, it is necessary to design a DNA circuit specialized for a specific circuit function. We aimed to achieve “plasticity” of the DNA circuit by addressing the motivating example. Our method was verified by numerical simulations based on reaction kinetics while utilizing dependable computer-aided software. |
|
|
Differential Dynamic Programming Approach for the Closed-loop Operation in Parameter-dependent System |
|
Hyein Jung(Seoul National University, Korea), Jong Woo Kim(Technische Universität Berlin, Germany), Jong Min Lee(Seoul National University, Korea) |
| A differential dynamic programming (DDP) approach was applied to adaptive control in parameter-dependent system. The DDP approach offers a simple state feedback control policy by using first and second derivatives of the system dynamics and objective function. To implement it, a hyper-state was defined as an augmentation of state and parameter and used. A continuously stirred tank reactor system with time-varying feed inlet conditions and set-point value is considered as an illustrative simulation example. The approach outperformed the conventional one, and found a new steady state, avoiding constraint violation. |
|
|
Latent Variable Space Model Predictive Controller Design with Model Cofidence Constraints |
|
Dong Hwi Jeong(University of Ulsan, Korea) |
| In order to control the the time varying set-point or a reference trajectory, latent variable space model based based predictive controllers have been suggested using the historical batch data. However, the conventional methods have not considered the confidence of the predictive model and just utilized it for constructing the controller. To overcome this shortcoming, two latent variable space model predictive controllers with a constraint about model confidence is proposed in this study. |
|
|
Product Description Recipe Generation from 3D STEP Model for Autonomous Task Planning |
|
Hyonyoung Han(ETRI, Korea) |
| The proposed system analyzes the product, part and assembly step descriptions from the STEP model. The system consists of five modules. The step analyzer extracts the BOM from a STEP file, and configures components of each part separately. The part analyzer analyzes feature size and relative position from the reference point of each part. The image analyzer generates representative images of each part and analyzes the shape type of part. The task analyzer finds the pairs of parts that make up to the joint, and determines the assembly sequence of the joints. The recipe generator creates product recipe with descriptions of parts and tasks, and generates assembly instructions for the product. |
|