Unsupervised Path Regression Networks
This work addresses the problem of efficient and scalable shortest path planning for robotics and autonomous systems, providing a competitive alternative to supervised methods.
This paper tackles shortest path problems by directly regressing splines from a neural network. The method, trained unsupervisedly, outperforms state-of-the-art supervised learning baselines, offering a more scalable training pipeline and faster inference.
We demonstrate that challenging shortest path problems can be solved via direct spline regression from a neural network, trained in an unsupervised manner (i.e. without requiring ground truth optimal paths for training). To achieve this, we derive a geometry-dependent optimal cost function whose minima guarantees collision-free solutions. Our method beats state-of-the-art supervised learning baselines for shortest path planning, with a much more scalable training pipeline, and a significant speedup in inference time.