Earticle

최근 로봇공학은 규칙 기반 제어에서 범용 AI 기반 제어로 전환되고 있다. VLA(Vision-Language-Action) 모델은 시 각 정보와 자연어 명령을 이해해 행동을 생성하는 엔드투엔드 멀티모달 학습을 가능하게 하며, 차세대 물리적 인공지능 핵심 기 술로 주목받는다. 실제 적용을 위해서는 ROS와의 통합이 필수적이다. 본 논문은 RoboNeuron, RT-2, OpenVLA, CoT-VLA 등의 사례를 중심으로 시스템 구조와 데이터 흐름을 비교·분석하고, 물류·제조·서비스 분야 적용 사례를 검토했다. 또한 실시간성, 데이터 부족, 추론 오류, 통합 복잡성 등 기술적 과제를 도출하고, 모델 양자화, Sim-to-Real 전이, 자가 피드백 루프, 모듈형 브리지 구조 등을 대응 방안으로 제시하였다. 이를 통해 규칙 기반 로봇에서 VLA–ROS 기반 멀티모달 지능, 범용 물리적 인공지능으로 발전하는 기술·산업적 로드맵을 제시한다.

Recent robotics research is shifting from rule-based control to general-purpose AI-driven control. Vision-Language-Action (VLA) models enable end-to-end multimodal learning by interpreting visual inputs and natural language commands to generate actions and are emerging as a core technology for next-generation Physical AI. Integration with the Robot Operating System (ROS) is essential for real-time sensing, motor control, and simulation-based validation. This paper analyzes representative frameworks including RoboNeuron, RT-2, OpenVLA, and CoT-VLA, comparing system architectures and data flows, and reviews applications across logistics, manufacturing, and service robotics. Key challenges such as real-time constraints, data scarcity, inference errors, and integration complexity are identified, along with proposed solutions such as model quantization, Sim-to-Real transfer, self-feedback loops, and modular bridge structures. A roadmap from rule-based robots to VLA–ROS multimodal intelligence and general-purpose Physical AI is presented, offering technical and industrial insights.

In order to overcome the limitations on human accidents and reduce experimental costs during autonomous driving, studies using virtual simulator is conducted. In addition, by using open source-based frameworks and middleware, we develop a module that can be applied immediately to a real vehicle using a model developed for autonomous driving that has been studied in a virtual environment. In this paper, we use autonomous driving environment using a game engine-based CARLA simulator, and propose a model predictive control module by interlocking open-source ROS (Robot Operating System) and Autoware with the simulator. In addition, ROS manages vehicle information and sensor data and transmits them to the Autoware. In order to control autonomous driving, the control end part is implemented using MPC (Model Predictive Control) among Autoware APIs, which provides necessary functions such as 3D map generation and NDT matching algorithm for autonomous driving, location identification, object recognition, and vehicle control.

Purposes: This study was to validate the Robotic Smart Work System that can provides better working conditions and high productivity in unstructured environments like bio-industry, based on a tele-operation system for fruit harvesting with low cost 3-D positioning system on the laboratory level. Methods: For the Robotic Smart Work System for fruit harvesting and cultivation management in agriculture, a vision based tele-operating system and 3-D position information are key elements. This study proposed Robotic Smart Farming, an agricultural version of Robotic Smart Work System, and validated a 3-D position information system with a low cost omni camera and a laser marker system in the lab environment in order to get a vision based tele-operating system and 3-D position information. Results: The tasks like harvesting of the fixed target and cultivation management were accomplished even if there was a short time delay (30 ms ~ 100 ms). Although automatic conveyor works requiring accurate timing and positioning yield high productivity, the tele-operation with user's intuition will be more efficient in unstructured environments which require target selection and judgment. Conclusions: This system increased work efficiency and stability by considering ancillary intelligence as well as user's experience and knowhow. In addition, senior and female workers will operate the system easily because it can reduce labor and minimized user fatigue.

Polishing a die that has free-form surfaces is a time-consuming and tedious job, and requires a considerable amount of high-precision skill. In order to reduce the polishing time and cope with the shortage of skilled workers, an automatic polishing robot system was developed. The polishing robot system is composed of two subsystems, a three-axis machining center and a two-axis polishing head. The machining center is controlled by a FANUC controller, and the polishing head by DSP controller. The system has five degrees of freedom and is able to keep the polishing tool normal to the die surface during operation. To easily operate the developed polishing robot system, this study developed an integrated operating program in the Windows environment. The program consists of five modules: a polishing data generation module, a code separation module, a polishing module, a graphic simulator module, and a teaching module. Also, the automatic teaching system was developed to easily obtain teaching data and it consists of a three dimensional joystick and a proximity sensor. Also, to evaluate the performance of the integrated operating program and the polishing robot system, polishing experiments of a die of shadow mask were carried out.

According to the increasing interest and demand for the Autonomous Surface Vessels (ASV), the autonomous navigation system is being developed such as obstacle detection, avoidance, and path planning. In general, autonomous navigation algorithm controls the ship by detecting the obstacles with various sensors and planning path for collision avoidance. This study aims to construct and prove autonomous algorithm with integrated various sensor using the Robot Operating System (ROS). In this study, the safety zone technique was used to avoid obstacles. The safety zone was selected by an algorithm to determine an obstacle-free area using 2D LiDAR. Then, drift angle of the ship was controlled by the propulsion difference of the port and starboard side that based on PID control. The algorithm performance was verified by participating in the 2020 Korea Autonomous BOAT (KABOAT).

According to the increasing interest and demand for the Autonomous Surface Vessels (ASV), the autonomous navigation system is being developed such as obstacle detection, avoidance, and path planning. In general, autonomous navigation algorithm controls the ship by detecting the obstacles with various sensors and planning path for collision avoidance. This study aims to construct and prove autonomous algorithm with integrated various sensor using the Robot Operating System (ROS). In this study, the safety zone technique was used to avoid obstacles. The safety zone was selected by an algorithm to determine an obstacle-free area using 2D LiDAR. Then, drift angle of the ship was controlled by the propulsion difference of the port and starboard side that based on PID control. The algorithm performance was verified by participating in the 2020 Korea Autonomous BOAT (KABOAT).

We need general developing environment to control robot with effect but less energy. So, software and hardware tools are very important. In this paper, we present a general-purpose robot control language and its implementation on Real Time O/S and VME bus system. The system runs on the VMEexec Real Time Operating System and robot program is written in the "C" language. The developed program is linked with the robot control C library io produce an executable image. Under the developed robot control environment, the user can write a general high-level control program leaving all the specific information about the robot in a robot specific file.ific file.

The multi-task operating system called HRMTOS (HUNDAI Robot Multi-task Operating System) was developed for concurrent execution. HRMTOS consists of condition interpreter, queue constructor, task scheduler. Condition interpreter checks the status and condition of request, queue constructor makes queue according to the checked result by condition interpreter, and task scheduler finds the task that will be urgently executed by priority of queue after pending the current excuting task. HRMTOS could execute teaching, playback, monitoring function of multi-robot and could be used more effectively than other robot controllers.

Among the prerequisite abilities (perception of environment, path planning and position estimation) of an autonomous mobile robot, position estimation has been seldom studied by mobile robot researchers. In most cases, conventional positioin estimation has been performed by placing landmarks or giving the entrire environmental information in advance. Unlikely to the conventional ones, the study addresses a new method that the robot itself can select distinctive features in the environment and save them as landmarks without any a priori knowledge, which can maximize the autonomous behavior of the robot. First, an orjentaion probaility model is applied to construct a lcoal map of robot's surrounding. The feature of the object in the map is then extracted and the map is saved as landmark. Also, presented is the position estimation method that utilizes the correspondence between landmarks and current local map. In dong this, the uncertainty of the robot's current positioin is estimated in order to select the corresponding landmark stored in the previous steps. The usefulness of all these approaches are illustrated with the results porduced by a real robot equipped with ultrasonic sensors.