Chunwei Xing

In this project, we trained a RL policy to sample poses and joint configurations for a excavator to traverse different terrains given its global trajectory on the 2D map. Our method outperforms trditional uniform sampling methods and shows good generalization abilities.

We formulate a learning-based approach by constructing constellations from semantically meaningful objects and use Deep Graph Convolution Networks to map the constellation representation to a descriptor. Our method outperforms the PointNet and handcrafted constellation descriptors for global localization, and shows robustness against different types of noise.