arXiv:2605.21311v1 Announce Type: new Abstract: Modern vision systems can detect, track, and forecast urban actors at scale, yet translating perception outputs to urban design remains limited. We introduce DeCoR, a two-stage reinforcement learning framework that leverages flow observations to co-optimize crosswalk layout and network-level signal control. The design stage encodes the pedestrian network as a graph and learns a generative policy that parameterizes a Gaussian mixture model over crosswalk location and width, from which new crosswalks are sampled. For each layout, a shared control p
The proliferation of advanced vision systems and the maturing of reinforcement learning techniques enable novel AI applications in urban planning previously considered too complex.
This development indicates a tangible application of AI to solve complex, real-world urban infrastructure problems, suggesting a path to more efficient and safer cities.
Traditional, static urban design and traffic management approaches can now be dynamically co-optimized with AI, potentially leading to more adaptive and responsive urban environments.
- · Urban Planners
- · Smart City Technology Providers
- · Municipal Governments
- · Residents of Urban Areas
- · Traditional Traffic Engineering Consultancies (slow to adapt)
- · Inefficient Urban Transportation Systems
- · High-Emitting Vehicles (indirectly via optimized flow)
Urban traffic flow and pedestrian safety are improved through AI-driven co-optimization.
Demand for AI-powered urban infrastructure solutions increases, driving further investment in smart city technologies.
AI-optimized urban layouts become a standard, necessitating new regulatory frameworks for algorithm-driven urban development and potential ethical considerations regarding control.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG