Zero-Shot Trajectory Planning for Signal Temporal Logic Tasks
Ruijia Liu, Ancheng Hou, Xiao Yu, Xiang Yin
TL;DR
This work tackles planning trajectories that satisfy complex Signal Temporal Logic (STL) specifications under unknown dynamics. It introduces a three‑part, data‑driven framework that semantically decomposes STL tasks into time‑aware progresses, allocates feasible timed waypoints using a dynamics‑aware Time Predictor, and synthesizes trajectory chunks with diffusion models that are stitched into complete plans. The authors prove soundness of the decomposition and allocation steps, and validate the approach across Maze2D, Cube, and AntMaze, demonstrating robust long‑horizon task satisfaction with substantial planning speedups compared to diffusion baselines and performance comparable to or exceeding optimization‑based methods under limited dynamics information. The proposed zero‑shot generalization to unseen STL tasks relies solely on task‑agnostic offline trajectories, enabling scalable planning in uncertain environments. This framework advances practical STL planning for real robots by coupling logic‑level guarantees with data‑driven, scalable trajectory synthesis and stitching.
Abstract
Signal Temporal Logic (STL) is a powerful specification language for describing complex temporal behaviors of continuous signals, making it well-suited for high-level robotic task descriptions. However, generating executable plans for STL tasks is challenging, as it requires consideration of the coupling between the task specification and the system dynamics. Existing approaches either follow a model-based setting that explicitly requires knowledge of the system dynamics or adopt a task-oriented data-driven approach to learn plans for specific tasks. In this work, we address the problem of generating executable STL plans for systems with unknown dynamics. We propose a hierarchical planning framework that enables zero-shot generalization to new STL tasks by leveraging only task-agnostic trajectory data during offline training. The framework consists of three key components: (i) decomposing the STL specification into several progresses and time constraints, (ii) searching for timed waypoints that satisfy all progresses under time constraints, and (iii) generating trajectory segments using a pre-trained diffusion model and stitching them into complete trajectories. We formally prove that our method guarantees STL satisfaction, and simulation results demonstrate its effectiveness in generating dynamically feasible trajectories across diverse long-horizon STL tasks.