Trajectory Planning without Trajectory Data: A Manifold-Guided Approach

2026-10-05

AI Summary

Traditional trajectory planning relies on costly expert demonstrations and struggles to generalize to unseen constraints. This work proposes an unsupervised path planning method that eliminates the need for trajectory supervision by learning the geometric structure of the latent manifold solely from state observations to construct feasible paths. By integrating manifold learning with geometric modeling of the state space, the proposed approach generates high-quality trajectories without any trajectory-level data. Extensive evaluations on maze navigation and dual-arm robotic benchmarks demonstrate that the method achieves performance comparable to classical planners while significantly enhancing generalization to novel start-goal configurations and previously unseen constraints.

Abstract

A common way for trajectory planning is to leverage generative models trained on large collections of expert trajectories. At inference time, the model generates executable trajectories by conditioning on task goal constraints. However, trajectory-based methods rely on costly supervision, scale poorly with sequence length, and often generalize poorly to unseen constraints such as novel start-goal pairs. We propose an alternative to learn the underlying state-space manifold and use the geometry of the manifold for trajectory planning. This approach requires only state observations and enables generalization to unseen constraints by con- structing trajectories on the learned manifold of the state space. Experiments on Maze2D and robotic motion-planning benchmarks show that Ariadne constructs feasible paths from state-only supervision and generalizes to unseen start-goal combinations. On high-dimensional dual-arm planning, it remains competitive with trajectory-supervised and classical planners, while requiring no trajectory data for training.

Problem

• trajectory planning
• state-space manifold
• generalization
• state-only supervision

Innovation

• Manifold Learning
• Trajectory Planning
• State-only Supervision
• Generalization
• Motion Planning