AgentScapes-Designing Water Efficient Landscapes Using Distributed Agent-based Optimization

摘要

In arid and semi-arid regions, landscaping can reduce the energy use of a home and generate a more pleasing environment for the home's residents. However, water in these regions is a scarce resource, which makes landscaping a tradeoff between conflicting objectives-maximizing the growth on the landscape while minimizing the water use. This paper presents a new system, called AgentScapes, that uses agent-based modelling and distributed optimization to design water-conserving residential landscapes. The agent-based model included in AgentScapes combines the basic ideas of local search and the specifics of how plant communities emerge based on individual plant-plant interactions and responses to resources available on the landscape. Agents in this model are plants with different light and water requirements placed on a simulated landscape with specified light and water conditions. Individual plant agents in desirable locations survive, while struggling agents die off and re-emerge in a new location. These interactions enable each plant agent to search for a location that maximizes its growth and minimizes its water use. The agent-based model generates solutions that are within 95 percent of optimal in a fraction of the time required for exhaustive search solutions. In addition, the agent model is able to reproduce heuristics for low-water garden design and present a new option for reducing water use.