This paper is a report on Open Space Discussion (OSD) held in Evolutionary Computation Symposium 2021. The purpose of OSD is to share and discuss problems at hand and future research targets related to evolutionary computation. Discussion topics are voluntarily proposed by some of the participants, and other participants freely choose one to join in the discussion. Through free discussions based on the open space technology framework, it is expected that participants will have new research ideas and start some collaborations. This paper gives the concept of OSD and introduces six topics discussed this year. This paper also shows the responses to the questionnaire on OSD for future discussions, collaborations, and related events.
A new approach to extract the optimal speed-control strategy for air traffic controllers (ATCOs) is proposed. Air traffic demand is expected to grow in the next decades, causing the overcapacity of large-scale airports. Extended Arrival Management (E-AMAN) is becoming one of the potential operational concepts to accommodate this increasing demand. In E-AMAN, upstream ATCOs instruct pilots to increase or decrease cruise speed to effectively reduce the expected delay times and fuel consumption around large-scale airports. The E-AMAN system is designed according to the target area because the optimal speed-control depends on the characteristics of the target airport, airspace, and air traffic flow. Therefore, this study establishes a method to analyze the characteristics contributing to speed control. First, a rule-based simulator and multi-objective optimization are combined to search for the optimal speed for each aircraft. The deceleration speed instructed at the 150NM from Tokyo International Airport is used as the design variable to minimize the flight time of both takeoff-inflow and cruise-inflow flights. Finally, the decision trees are constructed for the two major route-clusters by using the obtained non-dominated solutions and 18 local information obtained during the speed control. As the results, the non-dominated solutions significantly reducing the both flight times are obtained. The decision trees clarify the features and their thresholds which contribute to the decision-making of speed control for each route-cluster. The results imply that the optimal speed-control strategy could vary depending on the airport and airspace. This research will contribute to expanding the understanding of common points and differences in the speed control by ATCOs with different airspaces.
In recent years, evidence-based policy-making (EBPM) has been called for to accommodate diverse stakeholders when local governments formulate new policies. Social simulation allows virtual observation of changes in social conditions resulting from various alternatives in policy-making. However, there has not been a generic social simulation for designing subsidy payment policies that can be used in various situations. The Evolutionary Computation Competition 2021 (EC Comp 2021), an optimization competition that has been held since 2017 and intends to promote interaction between industry and academia, asked participants to design subsidy payment policies with social simulation. EC Comp 2021 newly formulates a generic social simulation framework for designing subsidy payment policies. This social simulation estimates the effects of subsidy payment policies in response to changes in household economic conditions based on economic shock scenarios using statistically valid data on the residents in a city. This paper gives a detailed explanation of the subsidy payment design problem with the social simulation in EC Comp2021. This paper explains the participants’ optimization methods and their results, accompanied by a brief analysis of their results, and discusses the characteristics of the optimization problem.
This paper proposed a Markov chain-based local optima network (LON), representing search transitions of an evolutionary algorithm (EA). LON is a graph constructed by local optima as nodes and search transitions of an optimization algorithm as edges. This paper focused on a mutation-based (1+1)-EA as the target optimization algorithm and constructed its LON, which could estimate the success ratio to find the optimal solution and the time to reach it based on the Markov chain model. We generated the proposed LONs on NK-landscape problems with twenty variables and the different number of co-variables from two to five and discussed the relations among the success ratio, the convergence time, and quantitative features observed from the generated LONs. The results revealed that the optimal solution’s funnel ratio in the variable space greatly impacts the success ratio. Also, we showed that the estimation accuracy of the success ratio and the convergence time of the (1+1)-EA by the proposed LON increase as the number of objective function evaluations increases. The coefficient of determination of the success ratio prediction exceeded 0.9 when the number of objective function evaluations got more than one thousand.
This paper addressed a real-world item stock allocation optimization in ASKUL Corporation, which conducted the electric commerce business covering all over Japan. This work treated a thousand items and eight warehouses. The item stock allocation optimization problem was a combinatorial optimization problem, which determined whether each item should be stocked in each warehouse or not. The problem had multiple constraints, such as the capacities of warehouses and two objectives: the shipping cost and the average number of warehouse stocks. The constraint and objective functions executed an existing complex shipping system internally and should be deemed as a black box. For the black box problem with multiple constraints and objectives, this paper employed CNSGA-II, a representative evolutionary algorithm, and a neighborhood cultivation mechanism. Since the conventional uniform crossover was too destructive for this problem, this paper proposed semantic crossovers grouping and crossing variables in units of item or warehouse. Experiments used real-world data, and results showed that the item crossover crossing each itemunit variable is appropriate for the item stock allocation optimization problem. Also, results showed that the obtained item stock allocation plans are better than an actually used human-made plan in both viewpoints of the shipping cost and the average number of warehouse stocks.