Kine2Go: Kinematic dataset for the Unitree Go2 robot with diverse gaits and motions

2026-06-12

AI Summary

This study addresses the lack of high-quality, diverse kinematic datasets with motor-level action labels for the Unitree Go2 quadruped robot, which has hindered the application of data-driven approaches such as imitation and reinforcement learning. To overcome this limitation, the authors propose an efficient data generation pipeline that combines cross-morphology gait transfer with reinforcement learning. Specifically, they transfer a variety of quadrupedal gaits onto the Go2 platform and train robust tracking policies to collect kinematic trajectories along with corresponding motor actions under perturbations. The resulting Kine2Go dataset comprises 40 distinct policies and 800 trajectories, offering the first large-scale, diverse, and motor-annotated motion dataset for Go2. This resource significantly reduces the cost of acquiring demonstration data and provides strong support for research and development in data-driven robotic control.

Abstract

The recent popularity of robotics, combined with the steadily decreasing cost of robotic hardware, has lowered the entry barrier to robotics research and enabled rapid advancements in the field. One of the primary examples is the Unitree Go2 quadruped robot, which is often used by researchers in the areas of locomotion, navigation, control, and others. Many researchers use the Go2 robot in combination with techniques like imitation learning, reinforcement learning, and behavioral cloning to allow machine learning systems to take full control of the robot. At the same time, many of those techniques require demonstration data consisting of the robot's kinematics information and actions applied to the motors. Obtaining such data is difficult, requires building complex pipelines, and can take significant time. To aid in those kinds of efforts, we present Kine2Go - a dataset with 800 diverse gait kinematics trajectory motion data for the Unitree Go2 robot, derived from 40 distinct policies. Our pipeline accepts data from various quadruped morphologies and translates them to a Go2-compatible format. Then we use Reinforcement Learning to train policies following a given motion, and finally we gather data from those policies, which grants robust, perturbed kinematic data with corresponding motor-level actions.

Problem

• kinematic dataset
• Unitree Go2
• quadruped robot
• gait data
• motion trajectories

Innovation

• kinematic dataset
• quadruped robot
• motion retargeting
• reinforcement learning
• imitation learning