[2I3-GS-5b] エージェント：マルチエージェントシミュレーション

[2I3-GS-5b-03] Guidance Effect Estimation on Crowd Movements Using Causal Inference

〇Koh Takeuchi^1,4,5, Ryo Nishida², Hisashi Kashima^1,4, Masaki Onishi³ (1. Kyoto University, 2. Tohoku University, 3. AIST AIRC, 4. RIKEN AIP, 5. JST PRESTO)

Keywords:Deep Learning, Causal Inference, Multi-Agent Simulator

Easing Traffic congestion in unusual events and evacuating people from an emergency-affected area are serious problems in our society. Guiding crowd movement is one of the popular candidates to deal with these problems. Hence, there has been demand for a decision support system that can answer a typical question: ``what will be the outcomes of each of the possible options in the current situation?". We consider the problem of estimating the effects of crowd movement guidance from past data. Since the amount available data is limited and biased by past decision-makers, we leverage two recently advanced techniques in deep representation learning for spatial data analysis and causal inference. We use a multi-agent simulator to generate realistic data on evacuation scenarios in a crowded theater to assess the performance on estimating the treatment effects of possible guidance. The results of experiments show that our method reduces the estimation error by at most 56\% from state-of-the-art methods.