MEGUMI FUJITA   Japanese

e-mail　
Affiliated post　National Institute of Informatics
The post　Principles of Informatics Research Division
Job name　Project researcher

Research Field

Intelligent robot control

Society Activity

Member of The Robotics Society of Japan
JAWS(Joint Agent Workshop and Symposium) Program Committee member 2012-2016
ICA(International Conference on Agents), Program Committee member 2016-

Introduction of my Research and Movies

1. My early study
   I made a robot which can explore the maze composed of grid map.
   
   
   Robot can look around while moving.
   Robot can make its belief by the information of its perception.
   
   
   Robot made wrong belief by misinformation, but corrected its belief by changing its search strategy made by dynamic planner.
   
   
   Since the end of the map is the robot's destination, the robot has returned to its initial position.
   
   
   I implemented BDI agent for robot control, and the robot is operated by remote control from the PC.
   
   
   
   
   
2. The reseach I did during my doctoral course Part.1
   I made the following two robots: the explorer which can search around in the map and it can detect a stand to put goods on, and the lift which can carry the goods. They can take cooperative actions by autonomously.
   
   
   When the lift finds an obstacle, since the lift can not distinguish an obstacle whether it is a stand or not, it delegates a task to check whether it is actually a stand to the explorer.
   
   
   In this experiment, the location of the robot was largely departed from its ideal location after moving, so the location of the robot had to be manually corrected.
   
   
   
   
   
3. The reseach I did during my doctoral course Part.2
   This experiment is another version of the above.
   
   
   When the explorer is near the stand, it found the stand, then consign the lift to put the goods.
   
   
   Where, the robots communicate each present position so as not to collide each other.
   
   
   
   
   
4. The research of the robot which can move autonomously in real world
   The robot can determine its action after looking around, and using this perception the robot's atomic actions are interlocked with the robot's movement in real world.
   
   
   The robot doesn't have the environmental map. Its perception and atomic action which includes reactive action makes adapt the robot's actual action and position in real world.
   
   
   I gave the approximate direction of the target for the robot's initial belief.
   
   
   I implemented this control program using Prolog, and the robot is operated remotely from the PC.
   
   
   
   
5. The research for the robot to move out of the room
   The basic implementation has been the continuation of the previous study.
   
   
   I implemented BDI agent for controling this robot.
   
   
   The achivement goal is "move out of the room", then the decomposed subgoals are "go to flont of the door" and "go to exit of the room".
   
   
   I let the robot learn the charcteristics of the door with SVM, and I gave it the empirical rule on door angle and entrance angle.
   
   
   
   

   Article

   Toward a robot that acquires logical recognition of space. Megumi Fujita, Yuki Goto, Naoyuki Nide, Ken Satoh, and Hiroshi Hosobe. Information Engineering Express, Vol. 3, No. 4, pp. 1-10, Dec, 2017.

   Implementation of 3-valued paraconsistent logic programming towards decision making system of agents. Yuki Goto, Megumi Fujita, and Naoyuki Nide. Journal of Systems Science and Systems Engineering, pp. 1-18, May, 2018. DOI: 10.1007/s11518-018-5367-7.

   BDI Robots Who Adapt to the Diversity of the Real World. Megumi Fujita and Hiroko Katayama and Naoyuki Nide and Shiro Takata IPSJ Transactions on MPS, vol. 5, No. 1, pp. 50-64. 2012(In Japanese)

   International Conference

   Autonomous Control of Mobile Robots Using Logical Representation of Map and Inference of Location. Megumi Fujita, Yuki Goto, Naoyuki Nide, Ken Satoh, Hiroshi Hosobe. IEEEICA 2016, 2016.

   Logic-based and robust decision making for robots in real world. Megumi Fujita, Yuki Goto, Naoyuki Nide, Ken Satoh, Hiroshi Hosobe. AAMAS2014 Demo session, pp.1685-1686 , 2014.

   An architecture for autonomously controlling robot with embodiment in real world. Megumi Fujita, Yuki Goto, Naoyuki Nide, Ken Satoh, Hiroshi Hosobe. Knowledge Representation and Reasoning in Robotics, Workshop at ICLP2013, 2013.

   Importing dynamic planner to BDI agent creating flexible decision-making of policies for selecting robot actions in real world. Megumi Fujita, Hiroko Katayama, Yuko Ojima, and Naoyuki Nide. In International Symposium on Artificial Life and Robotics, 2012.

   BDI logic with probabilistic transition and fixed-point operator. Naoyuki NIDE, Shiro Takata, and Megumi Fujita. In Proc. of CLIMA ’09, pp. 71–86, 2009.

   Doctoral Thesis

   Extension and application of behavior decision algorithm of autonomous agent based on BDI model. Megumi Fujita. Graduate School of Nara Women's Unviersity, pp. 1-63, 2012(In Japanese).

   Jun 4, 2018 updated