A METHOD FOR ROAD NETWORK GENERATION BASED ON TENSOR FIELD AND MULTI-AGENT

Abstract. Urban form generative design and optimization, through searching the whole design solution space efficiently to approximate the optimal ones, is a powerful means for supporting energy-efficient urban design. In this whole workflow, urban form generation is of great significance. It provides prototypes for the following steps of urban performance evaluation and optimization, and receives feedback to refine urban forms iteratively. For top-down urban form generation methods, road network generation is a prerequisite for the following generation of blocks, buildings, etc. This paper presented a method for generating hierarchical road networks based on tensor field and multi-agents. Compared to traditional generation methods based on tensor field, this approach introduced road planning indices, including density of road network, distance between road intersections, etc., to constrain the movement of agents to make the networks more reasonable. According to the method, a prototype of a tool for road network generation was developed based on ArcGIS Engine. Generation experiments were conducted through the developed tool. And the generated road networks can well reflect the impact of the geometry of the on-site elements. What’s more, the desired density of road networks and the requirement about distance between road intersections can be well satisfied.