Identifying adaptive pathways by surrogating computationally expensive agent-based models

摘要: Planning future cities to embrace new technologies such as automated, connected and electric vehicles is challenging due to complex interactions between mobility, infrastructure and land use. Furthermore, the timing for technologies to be ready, to what extent they are going to be adopted and how people interact with such technologies are highly uncertain. In this regard, the extensive study of current and future mobility changes and their potential impact on people’s behaviour and society allows an improved understanding of urban systems. However, despite the increasing recognition that cities face deep uncertainty, traditional spatial planning tools are inadequate to deal with it. Most research efforts have focused on increasing the granularity and complexity of models to gain accuracy, which further impedes a wide exploration of future scenarios. In this work, we explore the potential of using complex models that are capable of capturing the spatial interactions between mobility, infrastructure and land use, such as the agent-based MATSim, under deep uncertainty. We study potential urban measures (network configurations, Pick-Up/Drop-Off points, parking strategies and intersection designs) in response to the technological shift in transportation using the Tanjong Pagar fictive neighbourhood, in Singapore, as the test site. The model was evaluated under three main uncertainties over time: travel demand, automation development and vehicle-sharing preferences and then fit with a surrogate model. The surrogate was then used to search for optimal sequencing, timing and grouping of interventions (i.e. adaptive pathways) to provide guidance …