Existing navigation methods are primarily designed for specific robot embodiments, limiting their generalizability across diverse robot platforms. In this paper, we introduce X-Nav, a novel framework for end-to-end cross-embodiment navigation where a single unified policy can be deployed across various embodiments for both wheeled and quadrupedal robots. X-Nav consists of two learning stages: 1) multiple expert policies are trained using deep reinforcement learning with privileged observations using a wide range of randomly generated robot embodiments; and 2) a single general policy is distilled from the expert policies via navigation action chunking with transformer (Nav-ACT). The general policy directly maps visual and proprioceptive observations to low-level control commands, enabling generalization to novel robot embodiments. Simulated experiments demonstrated that X-Nav can effectively achieve zero-shot transfer to both unseen embodiments and photorealistic environments. A scalability study showed that the performance of X-Nav scales with the increasing number of randomly generated embodiments used during training. An ablation study confirmed the design choices of X-Nav. Furthermore, real-world experiments were conducted to validate the generalizability of X-Nav in real-world environments.
The proposed X-Nav framework consists of two stages: 1) Expert Policy Learning, and 2) General Policy Distillation. In Stage 1, multiple expert policies are trained on randomly generated robot embodiments using deep reinforcement learning with privileged observations. In Stage 2, the knowledge of expert policies is distilled into a single general policy using imitation learning.
