Generation of Paths in a Maze using a Deep Network without Learning
This addresses path planning in robotics or navigation by offering a highly efficient, training-free method for large-scale mazes, though it appears incremental as it builds on existing network architectures without learning.
The paper tackled the problem of path planning for multiple start- and end-points in large mazes by using a deep network with only max pooling layers that requires no training, achieving efficient solutions for mazes with over half a billion nodes and thousands of endpoints in short time on parallel hardware.
Trajectory- or path-planning is a fundamental issue in a wide variety of applications. Here we show that it is possible to solve path planning for multiple start- and end-points highly efficiently with a network that consists only of max pooling layers, for which no network training is needed. Different from competing approaches, very large mazes containing more than half a billion nodes with dense obstacle configuration and several thousand path end-points can this way be solved in very short time on parallel hardware.