• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1525-1530
• /
• 2003

In this paper, a model of adaptive individual life support system is proposed from the viewpoint of cybernetics. This model is derived based on the relation between human behavior and human action, static and dynamic in processing speed, and abstract/concrete. In applications, task and information of human which includes in this system analyzed by paying attention to cybernetics. This paper shows a few actual example of modeling by fundamental adaptive individual life support model such as medical diagnosis, health care and education support. Finally as an example, design and implementation are concretely carried out for health care support system. This is also a method to design a information support system which is involved in human.

• Journal of Mechanical Science and Technology
• /
• v.19 no.9
• /
• pp.1742-1752
• /
• 2005

This paper presents a Multiple Target Tracking (MTT) Adaptive Cruise Control (ACC) system which consists of three parts; a multi-model-based multi-target state estimator, a primary vehicular target determination algorithm, and a single-target adaptive cruise control algorithm. Three motion models, which are validated using simulated and experimental data, are adopted to distinguish large lateral motions from longitudinally excited motions. The improvement in the state estimation performance when using three models is verified in target tracking simulations. However, the performance and safety benefits of a multi-model-based MTT-ACC system is investigated via simulations using real driving radar sensor data. The MTT-ACC system is tested under lane changing situations to examine how much the system performance is improved when multiple models are incorporated. Simulation results show system response that is more realistic and reflective of actual human driving behavior.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.29 no.1 s.232
• /
• pp.139-144
• /
• 2005

A vehicle cruise control algorithm using an Interacting Multiple Model (IMM)-based Multi-Target Tracking (MTT) method has been presented in this paper. The vehicle cruise control algorithm consists of three parts; track estimator using IMM-Probabilistic Data Association Filter (PDAF), a primary target vehicle determination algorithm and a single-target adaptive cruise control algorithm. Three motion models; uniform motion, lane-change motion and acceleration motion. have been adopted to distinguish large lateral motions from longitudinal motions. The models have been validated using simulated and experimental data. The improvement in the state estimation performance when using three models is verified in target tracking simulations. The performance and safety benefits of a multi-model-based MTT-ACC system is investigated via simulations using real driving radar sensor data. These simulations show system response that is more realistic and reflective of actual human driving behavior.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.931-934
• /
• 2005

This paper describes an adaptive recommendation system that provides real-time personalized trading advice to the investors based on their profiles and trading information environment. A proposed system integrates Stochastic technical analysis and artificial neural network that incorporates an adaptive user modeling. The user model is constructed and updated based on initial user profile and recorded user interactions with the system. The information presented to each individual user is also tailor-made to fit the user's behavior and preference. A system prototype was implemented in JAVA. Experiments used to evaluate the system's performance were done on both human subjects and synthetic users. The results show our proposed system is able to rapidly learn to provide appropriate advice to different types of users.

• Journal of the Korean Geographical Society
• /
• v.43 no.1
• /
• pp.104-116
• /
• 2008

Overheated speculation areas which have high potential of becoming speculative are the target of many real estate policies. This paper proposes a model for spatial patterns of house price volatility and suggests a spatial pattern of overheated speculation areas. House prices are determined by economic behaviors of sellers and buyers who have rational or adaptive expectations. Spatial patterns of house price volatility are formed by tendencies of their economic behavior. If there is a majority of adaptive sellers and buyers in an area, it may appear as a "hotspot" by showing high volatility of house prices and simultaneous price increases. Overheated speculation areas are formed by adaptive sellers and buyers who want to realize maximum expectation profit, therefore these areas patterns are defined as hotspot patterns of price volatility.

• Proceedings of the KSME Conference
• /
• 2004.11a
• /
• pp.696-701
• /
• 2004

A vehicle cruise control algorithm using an Interacting Multiple Model (IMM)-based Multi-Target Tracking (MTT) method has been presented in this paper. The vehicle cruise control algorithm consists of three parts; track estimator using IMM-Probabilistic Data Association Filter (PDAF), a primary target vehicle determination algorithm and a single-target adaptive cruise control algorithm. Three motion models; uniform motion, lane-change motion and acceleration motion, have been adopted to distinguish large lateral motions from longitudinal motions. The models have been validated using simulated and experimental data. The improvement in the state estimation performance when using three models is verified in target tracking simulations. The performance and safety benefits of a multi-model-based MTT-ACC system is investigated via simulations using real driving radar sensor data. These simulations show system response that is more realistic and reflective of actual human driving behavior.

• Journal of Korea Game Society
• /
• v.8 no.1
• /
• pp.79-90
• /
• 2008

This paper describes a motivation-based hierarchical behavior planning framework to allow autonomous agents to select adaptive actions in simulation game environments. The combined behavior planning system is formed by four levels of specification, which are motivation extraction, goal list generation, action list determination and optimization. Our model increases the complexity of virtual human behavior planning by adding motivation with sudden and cumulative attributes. The motivation selection by probability distribution allows NPC to make multiple decisions in certain situations in order to embody realistic virtual humans. Hierarchical goal tree enhances the effective reactivity. Optimizing for potential actions provides NPC with safe and satisfying actions to adapt to the virtual environment. A restaurant simulation game was used to elucidate the mechanism of the framework.

• Journal of Korean Society of Transportation
• /
• v.26 no.2
• /
• pp.121-133
• /
• 2008

Conventional car following models are controlled when the velocity of following vehicle is equal to preceding vehicle without consideration of relative distance. Also, since the car following models are hardly consider the driver's behaviors and the environmental factors in driving, the models can't be adopted in reality. Hence, we developed the car following model applying Human Factors to consider driver's safety and comfortness. We simulated to compare the suggested model with the existing model, GGM(General GM). As results of simulations, the GGM model followed the preceding vehicle when the velocity of following vehicle was equal to preceding vehicle without relation of relative range. The other side, when the relative range was less or over than safety range, the suggested model made the relative range equal to safety range. Accordingly, we could be sure that the model would decrease the driver's discomfort and intensify the safety on driving without unnecessary waste of road. We identified that the suggested model is more realistic than the conventional GGM model.

• Journal of the Korea Computer Graphics Society
• /
• v.15 no.3
• /
• pp.1-9
• /
• 2009

Traditional computer simulation uses only traditional input and output devices. With the recent emergence of haptic techniques, which can give users kinetic and tactile feedback, the field of computer simulation is diversifying. In particular, as the virtual-reality-based surgical simulation has been recognized as an effective training tool in medical education, the practical virtual simulation of surgery becomes a stimulating new research area. The surgical simulation framework should represent the realistic properties of human organ for the high immersion of a user interaction with a virtual object. The framework should make proper both haptic and visual feedback for high immersed virtual environment. However, one model may not be suitable to simulate both haptic and visual feedback because the perceptive channels of two feedbacks are different from each other and the system requirements are also different. Therefore, we separated two models to simulate haptic and visual feedback independently but at the same time. We propose an adaptive mass-spring method as a multi-modal simulation technique to synchronize those two separated models and present a framework for a dual model of simulation that can realistically simulate the behavior of the soft, pliable human body, along with haptic feedback from the user's interaction.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.