MENU

RESEARCH

Among technologies of so-called AI (artificial intelligence), our laboratory approaches machine learning, mathematical optimization, urban transportation, disaster mitigation, ground classification, railways (rescheduling, stopping control), and electricity interchange optimization.

Therefore, we preferentially accept students in the fourth area (urban information engineering), of the department of the Urban Environment Systems. Students in the first area (urban planning and design) who interested in emergent design by artificial life such as evolutionary computation, facility placement by multi-agent technology, approach by the autonomous distributed system, human decision-making such as affordance, are also welcome.

Introduction

In the past, "intelligent ___" became popular, but now AI, which emphasizes deep learning, is a fad.

Indeed the ability to recognize and discriminate may have surpassed humans by being supported by a large amount of data.

BUT what about "decision" and "action" based on recognition?

  • How can we use it even if we can create new energy or recognize the current situation?
  • How the car controlled by human drives among autonomous cars?

Can you do the right thing even if you can recognize it?

"Do the Right Thing" - is the title of the book written by Professor Stuart Russell, one of my masters.

I think it should be noted that the expression RIGHT is used, not OPTIMAL.

In engineering and strictly closed world, we aim to "define an objective function and maximize it." In engineering and strictly closed world, we aim to "define an objective function and maximize it." However, in the world with uncertainty and dynamic change, in practice, the action is output (decided) while the object is slightly changed according to the situation (it is free to raise the bar or to drop). Knowing the setting mechanism of objects is the motivation of my AI research...

Dr. Russell's Q & A

What happens when everyone is smart, including machines and humans?

  • Knowledge that maximizes their own interests
  • Knowledge that prioritizes the interests of their communities and their masters
  • Knowledge that hopes the restoration of the global environment

Which intellectual level should we say SMART?

Here comes the multi-agent dilemma problem.

In our laboratory, we are researching methods of "action" and "decision-making," which will be common ground for the benefit of the whole earth = system, including people using "traffic, railway, energy."

When thinking about the mechanism to think of both piece/individual and whole levels, "the emergence of cooperation" is likely to be the slogan, but in our laboratory, there are members who dislike a "peremptory bossy attitude." Our policy is that we want to minimize as much top-down binding as possible from the top.

"A bit of knowledge, flexibility, discretion according to circumstances"... I think it would be possible if we could afford to look at the neighborhood a little. The spread of social media and crowdsourcing in recent years is a wonderful product embodied in the trend of smartification of connecting "high-level knowledge" with "a bit of knowledge." These do not seem to be collaborations that were born from the way of thinking like "making the world better" by patriots of Meiji Restoration (so I like it)... do you want to shed light on such familiar phenomena one by one?

JAPANESE