Unsupervised Path Regression Networks

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.

Authors

• Michal Pándy
• Daniel Lenton
• Ronald Clark

Keywords

• Training
• Shortest path problem
• Supervised learning
• Pipelines
• Neural networks
• Cost function
• Manipulators
• Neural Network
• Shortest Path
• Path Planning
• Inference Time
• Optimal Path
• Optimal Cost Function
• Path Length
• Linear Interpolation
• Stochastic Gradient Descent
• Hard Constraints
• Planning Algorithm
• Inference Speed
• Planning Problem
• Fully-connected Network
• Gradient Step
• Good Initialization
• ResNet-50 Backbone
• Signed Distance Function
• Rapidly-exploring Random Tree
• Collision-free Path
• Straight-line Path
• Target Configuration
• RGB-D Images