研究者詳細 - 上野　史

Ensemble Heart Rate Extraction Method for Biological Data from Water Pressure Sensor. 査読

Fumito Uwano, Keiki Takadama

2018 AAAI Spring Symposia 2018年

Improving Sleep Stage Estimation Accuracy by Circadian Rhythm Extracted from a Low Frequency Component of Heart Rate. 査読

Akari Tobaru, Fumito Uwano, Takuya Iwase, Kazuma Matsumoto, Ryo Takano, Yusuke Tajima, Yuta Umenai, Keiki Takadama

2018 AAAI Spring Symposia 2018年

Strategy for Learning Cooperative Behavior with Local Information for Multi-agent Systems. 査読

Fumito Uwano, Keiki Takadama

PRIMA 2018: Principles and Practice of Multi-Agent Systems - 21st International Conference(PRIMA) 663 - 670 2018年

Recovery System Based on Exploration-Biased Genetic Algorithm for Stuck Rover in Planetary Exploration. 査読

Fumito Uwano, Yusuke Tajima, Akinori Murata, Keiki Takadama

Journal of Robotics and Mechatronics 29 ( 5 ) 877 - 886 2017年10月

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：Fuji Technology Press

Contributing toward continuous planetary surface exploration by a rover (i.e., a space probe), this paper proposes (1) an adaptive learning mechanism as the software system, based on an exploration-biased genetic algorithm (EGA), which intends to explore several behaviors, and (2) a recovery system as the hardware system, which helps a rover exit stuck areas, a kind of immobilized situation, by testing the explored behaviors. We develop a rover-type space probe, which has a stabilizer with two movable joints like an arm, and learns how to use them by employing EGA. To evaluate the effectiveness of the recovery system based on the EGA, the following two field experiments are conducted with the proposed rover: (i) a small field test, including a stuck area created artificially in a park
and (ii) a large field test, including several stuck areas in Black Rock Desert, USA, as an analog experiment for planetary exploration. The experimental results reveal the following implications: (1) the recovery system based on the EGA enables our rover to exit stuck areas by an appropriate sequence of motions of the two movable joints
and (2) the success rate of getting out of stuck areas is 95% during planetary exploration.

DOI： 10.20965/jrm.2017.p0877

Supporting the Exploration of the Learning Goals for a Continuous Learner Toward Creative Learning. 査読

Takato Okudo, Tomohiro Yamaguchi 0002, Akinori Murata, Takato Tatsumi, Fumito Uwano, Keiki Takadama

Journal of Advanced Computational Intelligence and Intelligent Informatics 21 ( 5 ) 907 - 916 2017年9月

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：Fuji Technology Press

This paper proposes a learning goal space that visualizes the distribution of the obtained solutions to support the exploration of the learning goals for a learner. Subsequently, we examine the method for assisting a learner to present the novelty of the obtained solution. We conduct a learning experiment using a continuous learning task to identify various solutions. To assign the subjects space to explore the learning goals, several parameters related to the success of the task are not instructed to the subjects. In the comparative experiment, three types of learning feedbacks provided to the subjects are compared. These are presenting the learning goal space with obtained solutions mapped on it, directly presenting the novelty of the obtained solutions mapped on it, and presenting some value that is slightly related to the obtained solution. In the experiments, the subjects to whom the learning goal space or novelty of the obtained solution is shown, continue to identify solutions according to their learning goals until the final stage in the experiment is attained. Therefore, in a continuous learning task, our supporting method of directly or indirectly presenting the novelty of the obtained solution through the learning goal space is effective.

DOI： 10.20965/jaciii.2017.p0907

Comparison Between Reinforcement Learning Methods with Different Goal Selections in Multi-Agent Cooperation. 査読

Fumito Uwano, Keiki Takadama

Journal of Advanced Computational Intelligence and Intelligent Informatics 21 ( 5 ) 917 - 929 2017年9月

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：Fuji Technology Press

This study discusses important factors for zero communication, multi-agent cooperation by comparing different modified reinforcement learning methods. The two learning methods used for comparison were assigned different goal selections for multi-agent cooperation tasks. The first method is called Profit Minimizing Reinforcement Learning (PMRL)
it forces agents to learn how to reach the farthest goal, and then the agent closest to the goal is directed to the goal. The second method is called Yielding Action Reinforcement Learning (YARL)
it forces agents to learn through a Q-learning process, and if the agents have a conflict, the agent that is closest to the goal learns to reach the next closest goal. To compare the two methods, we designed experiments by adjusting the following maze factors: (1) the location of the start point and goal
(2) the number of agents
and (3) the size of maze. The intensive simulations performed on the maze problem for the agent cooperation task revealed that the two methods successfully enabled the agents to exhibit cooperative behavior, even if the size of the maze and the number of agents change. The PMRL mechanism always enables the agents to learn cooperative behavior, whereas the YARL mechanism makes the agents learn cooperative behavior over a small number of learning iterations. In zero communication, multi-agent cooperation, it is important that only agents that have a conflict cooperate with each other.

DOI： 10.20965/jaciii.2017.p0917

The robust spacecraft location estimation algorithm toward the misdetection crater and the undetected crater in SLIM 査読

Haruyuki Ishii, Keiki Takadama, Akinori Murata, Fumito Uwano, Takato Tatsumi, Yuta Umenai, Kazuma Matsumoto, Hiroyuki Kamata, Takayuki Ishida, Seisuke Fukuda, Shujiro Sawai, Shinichiro Sakai

Proceedings of International Symposium on Space Technology and Science, ISTS 2017 2017年6月

Strategies to Improve Cuckoo Search Toward Adapting Randomly Changing Environment. 査読

Yuta Umenai, Fumito Uwano, Hiroyuki Sato, Keiki Takadama

Advances in Swarm Intelligence - 8th International Conference 10385 573 - 582 2017年

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）出版者・発行元：Springer

Cuckoo Search (CS) is the powerful optimization algorithm and has been researched recently. Cuckoo Search for Dynamic Environment (D-CS) has proposed and tested in dynamic environment with multi-modality and cyclically before. It was clear that has the hold capability and can find the optimal solutions in this environment. Although these experiments only provide the valuable results in this environment, D-CS not fully explored in dynamic environment with other dynamism. We investigate and discuss the find and hold capabilities of D-CS on dynamic environment with randomness. We employed the multi-modal dynamic function with randomness and applied D-CS into this environment. We compared D-CS with CS in terms of getting the better fitness. The experimental result shows the D-CS has the good hold capability on dynamic environment with randomness. Introducing the Local Solution Comparison strategy and Concurrent Solution Generating strategy help to get the hold and find capabilities on dynamic environment with randomness.

DOI： 10.1007/978-3-319-61824-1_62

Communication-Less Cooperative Q-Learning Agents in Maze Problem 査読

Fumito Uwano, Keiki Takadama

INTELLIGENT AND EVOLUTIONARY SYSTEMS, IES 2016 8 453 - 467 2017年

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）出版者・発行元：SPRINGER INT PUBLISHING AG

This paper introduces a reinforcement learning technique with an internal reward for a multi-agent cooperation task. The proposed method is an extension of Q-learning which changes the ordinary (external) reward to the internal reward for agent-cooperation under the condition of no communication. To increase the certainty of the proposed methods, we theoretically investigate what values should be set to select the goal for the cooperation among agents. In order to show the effectiveness of the proposed method, we conduct the intensive simulation on the maze problem for the agent-cooperation task, and confirm the following implications: (1) the proposed method successfully enable agents to acquire cooperative behaviors while a conventional method fails to always acquire such behaviors; (2) the cooperation among agents according to their internal rewards is achieved no communication; and (3) the condition for the cooperation among any number of agent is indicated.

DOI： 10.1007/978-3-319-49049-6_33

Reinforcement Learning with Internal Reward for Multi-Agent Cooperation: A Theoretical Approach. 査読

Adaptive Learning Based on Genetic Algorithm for Rover in Planetary Exploration 査読

Fumito Uwano, Akinori Murata, Keiki Takadama

Proceedings of The International Symposium on Artificial Intelligence, Robotics and Automation in Space, i-SAIRAS 2016 2016年6月

Fumito Uwano, Naoki Tatebe, Masaya Nakata, Keiki Takadama, Tim Kovacs

EAI Endorsed Transactions on Collaborative Computing 2 ( 8 ) e2 - 2 2016年5月

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：Association for Computing Machinery, Inc

This paper focuses on a multi-Agent cooperation which is generally difficult to be achieved without sufficient information of other agents, and proposes the reinforcement learning method that introduces an internal reward for a multi-Agent cooperation without sufficient information. To guarantee to achieve such a cooperation, this paper theoretically derives the condition of selecting appropriate actions by changing internal rewards given to the agents, and extends the reinforcement learning methods (Q-learning and Profit Sharing) to enable the agents to acquire the appropriate Q-values up- dated according to the derived condition. Concretely, the internal rewards change when the agents can only find better solution than the current one. The intensive simulations on the maze problems as one of test beds have revealed the following implications:(1) our proposed method successfully enables the agents to select their own appropriate cooperating actions which contribute to acquiring the minimum steps towards to their goals, while the conventional methods (i.e., Q-learning and Profit Sharing) cannot always acquire the minimum steps
and (2) the proposed method based on Profit Sharing provides the same good performance as the proposed method based on Q-learning.

DOI： 10.4108/eai.3-12-2015.2262878