Search papers, labs, and topics across Lattice.
Addressing the rigidity of traditional spatial optimization, CityPlanner casts real-world urban planning into an interactive, file-based execution sandbox where agents iteratively generate, evaluate, and refine spatial layouts. The framework makes long-horizon credit assignment tractable through an atomic-task reinforcement learning strategy that explicitly decouples initial plan generation from feedback-driven refinement. On real-world benchmarks, this sandbox-agent paradigm consistently outperforms both specialized, task-specific RL architectures and general-purpose LLM agents under practical cost and service constraints.
Urban spatial optimization can be solved far more effectively by framing it as an interactive code-and-eval sandbox with decomposed build-and-refine RL than by relying on rigid, task-specific RL solvers.
Urban planning is a real-world spatial optimization problem that requires selecting feasible actions from large candidate spaces under practical objectives such as cost and service quality. Existing optimization and reinforcement learning methods are effective for fixed formulations, but often depend on task-specific representations and constraint handling. We propose \emph{CityPlanner}, a sandbox-agent framework for executable urban planning. CityPlanner introduces \emph{UrbanSandbox}, a unified file-based environment where agents inspect task files, generate plans, run evaluators, and revise decisions based on executable feedback. To make learning tractable, we further propose atomic-task reinforcement learning, which decomposes long sandbox trajectories into \emph{BuildPlan} for initial construction and \emph{ImprovePlan} for feedback-based refinement. Experiments on a real-world benchmark show that CityPlanner consistently outperforms heuristic, task-specific RL, and general LLM-agent baselines. Ablations verify the contributions of UrbanSandbox, atomic-task RL, and iterative deployment. We release the code and dataset at https://anonymous.4open.science/r/co-agent-C1C8
