With increased human-robot interactions in industrial settings, a safe and reliable collision detection framework has become an indispensable element of collaborative robots. The conventional framework detects collisions by estimating collision monitoring signals with a particular type of observer, which is followed by collision decision processes. This results in unavoidable trade-off between sensitivity to collisions and robustness to false alarms. In this study, we propose a collision detection framework (CollisionNet) based on a deep learning approach. We designed a deep neural network model to learn robot collision signals and recognize any occurrence of a collision. This data-driven approach unifies feature extraction from high-dimensional signals and the decision processes. CollisionNet eliminates heuristic and cumbersome nature of the traditional decision processes, showing high detection performance and generalization capability in real time. We performed quantitative analysis and verified the performance of the proposed framework through various experiments.

Imaging flow cytometry (IFC) is an emerging technology that acquires single-cell images at high-throughput for analysis of a cell population. Rich information that comes from high sensitivity and spatial resolution of a single-cell microscopic image is beneficial for single-cell analysis in various biological applications. In this paper, we present a fast image-processing pipeline (R-MOD: Real-time Moving Object Detector) based on deep learning for high-throughput microscopy-based label-free IFC in a microfluidic chip. The R-MOD pipeline acquires all single-cell images of cells in flow, and identifies the acquired images as a real-time process with minimum hardware that consists of a microscope and a high-speed camera. Experiments show that R-MOD has the fast and reliable accuracy (500 fps and 93.3% mAP), and is expected to be used as a powerful tool for biomedical and clinical applications.

The image-activated cell sorter employs a significantly simplified operational procedure based on a syringe connected to a piezoelectric actuator and high-performance inference with TensorRT Integration.

Surface electromyography is widely used in many fields to infer human intention. However, conventional electrodes are not appropriate for long-term measurements and are easily influenced by the environment, so the range of applications of sEMG is limited. In this paper, we propose a flexible band-integrated, curved microneedle array electrode for robust long-term measurements, high selectivity, and easy applicability. Signal quality, in terms of long-term usability and sensitivity to perspiration, was investigated. Its motion-discriminating performance was also evaluated. The results show that the proposed electrode is robust to perspiration and can maintain a high-quality measuring ability for over 8 h. The proposed electrode also has high selectivity for motion compared with a commercial wet electrode and dry electrode.

Model based disturbance compensating methods such as disturbance observer, robust control with adaptive algorithm, enhanced internal model control are well known control structures for robust motion controller which can satisfy desired performance and robustness of high-speed/high-accuracy positioning systems. In this paper, these are analyzed in the unified framework and their design method is proposed. To do this, a generalized disturbance attenuation framework named RIC (robust internal-loop compensator) structure is introduced. Through parameterization based on RIC, it is shown that there are inherent equivalences in their structures and the proposed RIC gives a general design framework for the model based disturbance compensating methods. Through simulation, the proposed method is verified.

Most commercial optical sorting systems are designed to achieve high throughput, so they use a naive low-latency image processing for object identification. These naive low-latency algorithms have difficulty in accurately identifying objects with various shapes, textures, sizes, and colors, so the purity of sorted objects is degraded. Current deep learning technology enables robust image detection and classification, but its inference latency requires several milliseconds; thus, deep learning cannot be directly applied to such real-time high throughput applications. We therefore developed a super-high purity seed sorting system that uses a low-latency image-recognition based on a deep neural network and removes the seeds of noxious weeds from mixed seed product at high throughput with accuracy. The proposed system partitions the detection task into localization and classification, and applies batch inference only once strategy; it achieved 500-fps throughput image-recognition including detection and tracking. Based on the classified and tracked results, air ejectors expel the unwanted seeds. This proposed system eliminates almost the whole weeds with small loss of desired seeds, and is superior to current commercial optical sorting systems.