Identifying optimum bike station initial conditions using Markov chain modeling

Bike sharing systems (BSSs) are being deployed in many cities because of their environmental, social, and health benefits. To maintain low rental costs, rebalancing costs must be kept minimal. In this paper, we use BSS data collected from the San Francisco Bay Area to build a Markov chain model for each bike station. The models are then used to simulate the BSS to determine the optimal station-specific initial number of bikes for a typical day to ensure that the probability of the station becoming empty or full is minimal and hence minimizing the rebalancing cost