A5077919719- Distributed Control Multi-Agent Systems
- Robotic Path Planning Algorithms
- Target Tracking and Data Fusion in Sensor Networks
- Distributed Sensor Networks and Detection Algorithms
- Complex Network Analysis Techniques
- Advanced MIMO Systems Optimization
- UAV Applications and Optimization
- Cooperative Communication and Network Coding
- Misinformation and Its Impacts
- Opinion Dynamics and Social Influence
- Neural Networks and Applications
- Energy Harvesting in Wireless Networks
- Control and Dynamics of Mobile Robots
- Blind Source Separation Techniques
- Mobile Ad Hoc Networks
- Face and Expression Recognition
- Spectroscopy and Chemometric Analyses
- Insect Pheromone Research and Control
- Energy Efficient Wireless Sensor Networks
- Network Security and Intrusion Detection
- Opportunistic and Delay-Tolerant Networks
- Mobile Crowdsensing and Crowdsourcing
- Reinforcement Learning in Robotics
- Robotics and Sensor-Based Localization
- ICT Impact and Policies
Jet Propulsion Laboratory
2023-2024
California Institute of Technology
2023
Korea Advanced Institute of Science and Technology
2010-2021
Seoul National University
2021
University of Colorado Boulder
2016-2020
Korea Institute of Science & Technology Information
2020
University of Colorado System
2016-2019
University of Tennessee at Knoxville
2006-2015
Korea Air Force Academy
2012-2015
Soongsil University
2001-2003
As high-speed networks are becoming commonplace, it is increasingly challenging to prevent the attack attempts at edge of Internet. While many high-performance intrusion detection systems (IDSes) employ dedicated network processors or special memory meet demanding performance requirements, often increases cost and limits functional flexibility. In contrast, existing software-based IDS stacks fail achieve a high throughput despite modern hardware innovations such as multicore CPUs, manycore...
Many real-world reinforcement learning tasks require multiple agents to make sequential decisions under the agents' interaction, where well-coordinated actions among are crucial achieve target goal better at these tasks. One way accelerate coordination effect is enable communicate with each other in a distributed manner and behave as group. In this paper, we study practical scenario when (i) communication bandwidth limited (ii) share medium so that only restricted number of able...
This paper presents a novel block-based neural network (BBNN) model and the optimization of its structure weights based on genetic algorithm. The architecture BBNN consists 2D array fundamental blocks with four variable input/output nodes connection weights. Each block can have one different internal configurations depending settings, includes some restrictions such as integer in order to allow easier implementation reconfigurable hardware field programmable logic arrays (FPGA). are encoded...
Continually learning in the real world must overcome many challenges, among which noisy labels are a common and inevitable issue. In this work, we present replay-based continual framework that simultaneously addresses both catastrophic forgetting for first time. Our solution is based on two observations; (i) can be mitigated even with via self-supervised learning, (ii) purity of replay buffer crucial. Building regard, propose key components our method: technique named Self-Replay circumvent...
In this paper, we address the problem of associating mobile stations (MSs) with base (BSs) in an energy-efficient manner. We take a population game approach, which allows tractable analysis many selfish mobiles without growing mathematical complexity. From our game-theoretical analysis, prove that simple power-dependent pricing by operators leads Nash equilibrium to be equal optimal solution social optimization (i.e., no price-of-anarchy). study three evolution dynamics MSs, each expressed...
This paper presents a new hybrid dimensionality reduction method to seek projection through optimization of both structural risk (supervised criterion) and data independence (unsupervised criterion). Classification accuracy is used as metric evaluate the performance method. By minimizing risk, originated from decision boundaries directly improves classification supervised perspective. From an unsupervised perspective, can also be obtained based on maximum among features (or attributes) in...
Social networks are the major routes for most individuals to exchange their opinions about new products, social trends and political issues via interactions. It is often of significant importance figure out who initially diffuses information, i.e., finding a rumor source or trend setter. known that such task highly challenging detection probability cannot be beyond 31% regular trees, if we just estimate from given diffusion snapshot. In practice, entails process querying asks "Are you...
Recently, the problem of detecting rumor source in a social network has been much studied, where it shown that detection probability cannot be beyond 31% even for regular trees. In this paper, we study impact an anti-rumor on detection. We first show negative result: anti-rumor's diffusion does not increase under Maximum-Likelihood-Estimator (MLE) when number infected nodes are sufficiently large by passive starts to spread special node, called protector, after is reached rumor. next...
The carrier sense multiple access (CSMA) algorithm has been used in the wireless medium control (MAC) under standard 802.11 implementation due to its simplicity and generality. An extensive body of research on CSMA long made not only context practical protocols, but also a distributed way optimal MAC scheduling. However, current state-of-the-art (or extensions) still suffers from poor performance, especially multi-hop scenarios, often requires patch-based solutions rather than universal...
Although communication is a critical factor of situational awareness and performance on networked sUAS, most simulators do not provide realistic wireless modules. It prevents realizing among multiple UA information sharing, which yields significantly different results compared with actual in the real world. This paper presents communication-realistic simulator to improve more from coordination or cooperative planning/control sUAS. The this based Gazebo/ROS ArduPilot, where...
This paper presents block-based neural net works (BbNNs) for pattern classification. The BbNN achieves two goals: simultaneous optimization of network structure/weights; and implementation using reconfigurable digital hardware. BbNN, in a 2-D array basic blocks with four variable input/output nodes, successfully solve classification problems. Internal structure is optimized the GA encoding scheme.
This paper presents a framework for information-theoretic based task assignment of multiple UAVs searching and tracking moving targets. A stochastic approach which uses probability density functions is introduced as the information gathering metric. algorithm yields trajectory that minimizes uncertainty unknown information. The presented on negotiation activated when one defined conditions satisfied. can generate admissible solutions with minimal computational time. Cost computed by each...
In active source seeking, a robot takes repeated measurements in order to locate signal cluttered and unknown environment. A key component of an seeking planner is model that can produce estimates the at locations with uncertainty quantification. This allows plan for future Traditionally, this has been form Gaussian process, which difficulty scaling cannot represent obstacles. We propose global local factor graph scale large number obstacles combine extensions highly scalable system...
This paper presents two approaches considering a distributed framework for joint optimization of sensor coverage target detection and tracking maximizing estimation performance multi-agent systems. The first algorithm is based on the Lloyd algorithm, which uses centroid Voronoi partitions, one workarounds problems. other partition, result assigning to agent control input that maximizes information matrix corresponding state estimate target. Two presented algorithms are compared analyzed in...
This paper examines the problem of distributed path planning for a mobile sensor network comprised communication-aware robots performing general information gathering missions. Mutual is derived sensing over packet erasure channels that model multi-hop communication. We as non-cooperative game and derive utility functions are optimized locally by each robot. Each robot computes control input in manner results combined action can be bounded optimal centralized result utilizing sub-modularity...
In this paper, we study a problem of detecting the source diffused information by querying individuals, given sample snapshot diffusion graph, where two queries are asked: (i) whether respondent is or not, and (ii) if which neighbor spreads to respondent. We consider case when respondents may not always be truthful some cost taken for each query. Our goal quantify necessary sufficient budgets achieve detection probability 1- δ any 0 <; 1. To end, types algorithms: adaptive non-adaptive ones,...