Hanguen Kim

Cooperative Coevolutionary Algorithm-Based Model Predictive Control Guaranteeing Stability of Multirobot Formation

This paper proposes a novel cooperative coevolutionary algorithm (CCEA)-based distributed model predictive control (MPC) that guarantees asymptotic stability of multiagent systems whose state vectors are coupled and nonseparable in a cost function. While conventional evolutionary algorithm-based MPC approaches cannot guarantee stability, the proposed CCEA-based MPC approach guarantees asymptotic stability regardless of the optimality of the solution that the CCEA-based algorithm generates with a small number of individuals. To guarantee stability, a terminal state constraint is found, and then a repair algorithm is applied to all candidate solutions to meet the constraint. Furthermore, as the proposed CCEA-based algorithm finds the Nash-equilibrium state in a distributed way, robots can quickly move into a desired formation from their locations. A novel dynamic cooperatively coevolving particle swarm optimization (CCPSO), dynamic CCPSO (dCCPSO) in short, is proposed to deal with the formation control problem based on the conventional CCPSO, which was the most recently developed algorithm among CCEAs. Numerical simulations and experimental results demonstrate that the CCEA-based MPC greatly improves the performance of multirobot formation control compared with conventional particle swarm optimization-based MPC.

Weighted joint-based human behavior recognition algorithm using only depth information for low-cost intelligent video-surveillance system

Human joint estimation and behavior recognition algorithms are presented.Only depth information is used and can be executed on a low cost computing platform.The proposed system can be used with any subject instantly without pre-calibration.Experiments to verify the proposed algorithms have been conducted. Recent advances in 3D depth sensors have created many opportunities for security, surveillance, and entertainment. The 3D depth sensors provide more powerful monitoring systems for dangerous situations irrespective of lighting conditions in buildings or production facilities. To robustly recognize emergency actions or hazardous situations of workers at a production facility, we present human joint estimation and behavior recognition algorithms that solely use depth information in this paper. To estimate human joints on a low cost computing platform, we propose a human joint estimation algorithm that integrates a geodesic graph and a support vector machine (SVM). The human feature points are extracted within a range of geodesic distance from a geodesic graph. The geodesic graph is used for optimizing the estimation result. The SVM-based human joint estimator uses randomly selected human features to reduce computation. Body parts that typically involve many motions are then estimated by the geodesic distance value. The proposed algorithm can work for any human without calibration, and thus the system can be used with any subject immediately even with a low cost computing platform. In the case of the behavior recognition algorithm, the algorithm should have a simple behavior registration process, and it also should be robust to environmental changes. To meet these goals, we propose a template matching-based behavior recognition algorithm. Our method creates a behavior template set that consists of weighted human joint data with scale and rotation invariant properties. A single behavior template consists of the joint information that is estimated per frame. Additionally, we propose adaptive template rejection and a sliding window filter to prevent misrecognition between similar behaviors. The human joint estimation and behavior recognition algorithms are evaluated individually through several experiments and the performance is proven through a comparison with other algorithms. The experimental results show that our method performs well and is applicable in real environments.

Real-Time Human Pose Estimation and Gesture Recognition from Depth Images Using Superpixels and SVM Classifier

In this paper, we present human pose estimation and gesture recognition algorithms that use only depth information. The proposed methods are designed to be operated with only a CPU (central processing unit), so that the algorithm can be operated on a low-cost platform, such as an embedded board. The human pose estimation method is based on an SVM (support vector machine) and superpixels without prior knowledge of a human body model. In the gesture recognition method, gestures are recognized from the pose information of a human body. To recognize gestures regardless of motion speed, the proposed method utilizes the keyframe extraction method. Gesture recognition is performed by comparing input keyframes with keyframes in registered gestures. The gesture yielding the smallest comparison error is chosen as a recognized gesture. To prevent recognition of gestures when a person performs a gesture that is not registered, we derive the maximum allowable comparison errors by comparing each registered gesture with the other gestures. We evaluated our method using a dataset that we generated. The experiment results show that our method performs fairly well and is applicable in real environments.

Curvature Path Planning with High Resolution Graph for Unmanned Surface Vehicle

In this paper, we propose a curvature path planning algorithm for unmanned surface vehicles (USVs). To control the USV automatically, various robot navigation techniques can be applied and numerous researchers are working on a grid map-based path planning algorithms. However, the most grid map-based path planning methods for the USVs consider only two-dimensional (x, y) plane without considering vehicle’s maximum curvature. Since the most of the USVs are typically highly under-actuated, the ship tends to result in failure and sometimes induces hazardous collision situation when the ship follows the resultant path generated by the two-dimensional path planning algorithm. To solve this problem, we construct a non-uniform grid map which can reflect the geometric cost. Next we extend the dimension to reflect the kinematic constraint of the USV. Finally, to consider the vehicle’s dynamic constraint, we propose a new cost function. The result of the proposed algorithm has been demonstrated through the simulation on the real map and the results show that the proposed algorithm generates the most plausible and efficient path.

ViSP: visually servoed paired structured light system for measuring structural displacement

To inspect structural conditions, structural displacement is needed to be monitored at any time. Therefore, our previous study proposed a ViSP (Visually Servoed Paired structured light system) which is composed of two sides facing with each other, each with a camera, a screen, and one or two lasers controlled by a 2-DOF manipulator. In this system, the relative translational and rotational displacement between two sides can be estimated by calculating positions of the projected laser beams on the screens and the rotation angles of the manipulators. To validate the performance of the system, the various experimental tests with a two-story structural model were performed. The estimated results were compared with the results from a laser displacement sensor which can be considered as a reference. The results show that the presented system has potential of estimating the response of the structures with high accuracy in real time.

Gesture recognition algorithm for moving kinect sensor

Many studies have been conducted to achieve natural interaction between robots and humans, and this area of research is becoming more important. Except for several works, however, most gesture recognition algorithms are designed with an assumption that the sensor does not move. To resolve this problem, we propose an algorithm that can robustly recognize gestures while the sensor is moving.

Detection of a Suicide by Hanging Based on a 3-D Image Analysis

Prisoner suicide has been one of the leading causes of death in correctional facilities. In particular, since hanging is the most common suicide method, various systems have been proposed to detect and prevent hanging attempts. This letter presents the preliminary results of applying 3-D image recognition techniques to detect a partial suspension hanging attempt. This is the most frequently occurring method of hanging in prison settings, but is difficult to recognize using conventional methods. By using a 3-D camera and random forest, the experimental results show that the proposed detection algorithm can detect hanging suicide attempts with more than 95% accuracy.

Cooperative coevolution-based model predictive control for multi-robot formation

In this paper, a novel cooperative coevolution (CC)-based model predictive control (MPC) is proposed to deal with formation control problem. While conventional evolutionary algorithm (EA)-based MPC approaches cannot guarantee the stability, the proposed CC-based MPC guarantees the asymptotic stability regardless of the optimality of the solution that the CC-based algorithm generates with a small number of individuals. To guarantee the stability, a terminal state constraint is found, and then a repair algorithm is applied to all candidate solutions to meet the constraint. The cooperatively coevolving particle swarm optimization (CCPSO), most recently developed algorithm among CC-based EAs, is adopted and modified to fit into the formation control problem. Experiments are performed on a group of nonholonomic mobile robots to demonstrate the effectiveness of the CC-based MPC.

Remote guidance of untrained turtles by controlling voluntary instinct behavior.

Recently, several studies have been carried out on the direct control of behavior in insects and other lower animals in order to apply these behaviors to the performance of specialized tasks in an attempt to find more efficient means of carrying out these tasks than artificial intelligence agents. While most of the current methods cause involuntary behavior in animals by electronically stimulating the corresponding brain area or muscle, we show that, in turtles, it is also possible to control certain types of behavior, such as movement trajectory, by evoking an appropriate voluntary instinctive behavior. We have found that causing a particular behavior, such as obstacle avoidance, by providing a specific visual stimulus results in effective control of the turtles movement. We propose that this principle may be adapted and expanded into a general framework to control any animal behavior as an alternative to robotic probes.

Path Planning for Multi-agent Jellyfish Removal Robot System JEROS and Experimental Tests

Over the recent years, the increasing influence of climate change has given rise to an uncontrolled proliferation of jellyfish in marine habitats, which has visibly damaged many ecosystems, industries, and human health. To resolve this issue, our team developed a robotic system to successfully and efficiently remove jellyfishes, named JEROS (Jellyfish Elimination RObotic Swarm). The JEROS consists of multiple USVs (Unmanned Surface Vehicles) that freely move in a marine environment to scavenge for and eliminate jellyfishes. In this paper, we propose a constrained formation control algorithm that enhances the efficiency of jellyfish removal. Our formation control algorithm is designed in consideration of the characteristic features of JEROS. It is designed to effectively work with the simple leader-follower algorithm. The leader-follower formation control does not work well if a reference path of the leader is generated without considering a minimum turning radius. In order to overcome such a limitation, a new path planning method—angular rate-constrained path planning—is proposed in this paper. The performance of the jellyfish removal function was tested at Masan Bay in the Southern coast of South Korea and formation control tests were conducted at Bang-dong Reservoir in Daejeon, South Korea.