Q-Learning for Flexible Learning of Daily Activity Plans

摘要: Q-learning is a method from artificial intelligence to solve the reinforcement learning problem (RLP), defined as follows. An agent faced with set of states, S. For each state s there actions, A(s), that can take and takes (deterministically or stochastically) another state. receives (possibly stochastic) reward. The task select actions such reward maximized. Activity generation for demand in context transportation simulation. member synthetic population, daily activity plan stating sequence activities (e.g., home-work-shop-home), including locations times, needs be found. Activities at different generate transportation. modeled an RLP states given by triple (type activity, starting time already spent activity). possible are either stay move ...