We are interested in emergent phenomena caused by an interaction among agents who are implemented by computer programs.
We mainly focus on 2 topics;
(1) Multi-agent systems where agents cooperates with each other to solve given problems
(2) Social simulation where agents interacts according to their own goals
HTV cargo integration
Bus route network optimization
After The 2011 off the Pacific coast of Tohoku Earthquake, our lab tackles on the optimization of bus network at the time of disaster. The right side figure shows the example of the optimal bus network in Tokyo at the time of disaster. At the time of disaster, full suspension of the the train service as a main transportation network in Japan triggers five million commuters unable to get home. The bus network is focused on as the alternative means for transportation, since it is easy to change the bus route. Our lab proposes the multi-agent based route optimization methods. Now we tackle on the problem of a bus service called passenger traffic bottlenecks (i.e., the number of passengers exceeds usual demand which are accumulated by the number of stranded persons around bus
For safety and economic aircraft landing, it is required to optimize both the landing routes of multiple aircrafts and their landing sequence in real-time as the air transport service. Such landing route and landing sequence are important issue in the field of air traffic control science because air traffic controllers should determine both the landing routes and their landing sequence. This problem is called as the Aircraft Landing Problem (ALP). To tackle this issue, our lab proposes the optimization method that optimizes the solutions (from the viewpoint of the route distance) while promoting to increase the diversity of the solutions (from the viewpoint of a variety of the landing routes).
Sleep stage estimation method
Nursing care support
We research about agents who adapt to group, for example, AIBO released from SONY, PLEO and so on. Because of complexity of group, designing group-adapting agents is more difficult than individual one.
To solve those difficulty, we experiment with “BARNGA” from educational domain. BARNGA is a card game to experience different cultures. In addition, BARNGA can model real-world-groups easily. We have two steps to confirm the effect of BARNGA.
First, we play BARNGA with testee. By observing experiment, we try to define the state of group adaptation. Second, using results of subject experiment, we design agents who adapt to groups, and confirm the its effect by computer simulation. As you can see, we show a view of BARNGA simulation at right picture.
By results of simulation, we could get the following knowledge.
Learning classifier system