抄録
We propose a concept of human machine co-learning, in which the system learns the user's intention interactively, and the planner as a user learns the characteristics of the problem based on the solution represented by the system. In this paper, we propose a method to reflect the user's intention on the optimization by providing the interactive user interface connecting between "the design space", which is easier for a user to understand, and "the evaluation space", in which the system executes the optimization. We applied this idea to the NETSAT communication planning problem. NETSAT consists of existent multiple low-earth orbit satellites, and forms an ad hoc network dynamically according to the satellite position at the time when a certain communication is required. The effectiveness of our proposed method is evaluated by using the prototype system of the satellite simulator and interactive planner.
