Event
JEPLO proposes JEPA- and LiDAR-based quadruped gait method
1 reports · 1 sources
JEPLO proposes a terrain-adaptive gait control method for quadruped robots based on JEPA and LiDAR perception: after aggregating single-radar point clouds into depth observations, the network predicts the latent variables of the next timestep in latent space rather than reconstructing images, supplemented by teacher-student distillation and distribution regularization. The method claims that the pretrained weights have been reproduced in simulation; principal component analysis of the latent variables clusters clearly by terrain. Currently, public information is limited to simulation reproduction, and no validation results on real robots have been seen yet.
Report timeline
- JEPLO: Terrain-Adaptive Gait Control for Quadruped Robots Based on JEPA and LiDAR Perception
JEPLO proposes a terrain-adaptive gait control method for quadruped robots that aggregates single-LiDAR point clouds into depth observations and predicts the next-beat latent variable in latent space, without reconstructing images. The method is supplemented by teacher-student distillation and distribution regularization; pretrained weights are reproduced in simulation, and principal component analysis of the latent variables clusters clearly by terrain.
Guyuehome ·