Research Analyzer
← Back ICRA 2026

OmniNet: Omnidirectional Jumping Neural Network with Height-Awareness for Quadrupedal Robots

Yimin Han, Jiahui ZHANG, Zeren Luo, YINZHAO DONG, Jinghan Lin, Liu Zhao, Shihao Dong, Peng Lu

PDF

AI summary

Key figure (auto-extracted from paper)
A unified reinforcement learning framework enables quadrupedal robots to perform precise, omnidirectional jumps with stable landing without predefined trajectories or complex training curricula.
Legged Robots Reinforcement Learning Jumping Control Height Estimation Inverse Kinematics Agile Locomotion

Problem

Quadrupedal robots struggle to replicate agile, explosive jumping due to complex dynamics and limited aerial control, while existing methods rely on rigid trajectories, complex curricula, or separate estimators that hinder real-world versatility.

Approach

The authors train a jumping policy concurrently with an online height-aware state estimator and apply an analytical inverse kinematics reward to regulate aerial gestures, enabling direct tracking of commanded height and direction.

Key results

  • Precise tracking of user-specified jumping heights via integrated online state estimation
  • Omnidirectional jumping capability without predefined trajectories or reference datasets
  • Real-time aerial gesture regulation using an analytical inverse kinematics reward
  • Robust, seamless switching between jumping and walking gaits on uneven terrain in simulation and real hardware

Why it matters

Enables quadrupedal robots to execute versatile, agile jumps in unstructured environments, advancing dynamic locomotion for practical robotic applications.

Abstract

In the robotics community, it has been a long- standing challenge for quadrupeds to achieve highly explosive movements similar to their biological counterparts. In this work, we introduce a novel training framework that achieves height- aware and omnidirectional jumping for quadrupedal robots. To facilitate the precise tracking of the user-specified jumping height, our pipeline concurrently trains an estimator that infers the robot and its end-effector states in an online fashion. Besides, a novel reward is involved by solving the analytical inverse kinematics with pre-defined end-effector positions. Guided by this term, the robot is empowered to regulate its gestures during the aerial phase. In the comparative studies, we verify that this controller can not only achieve the longest relative forward jump distance, but also exhibit the most comprehensive jumping capabilities among all the existing jumping controllers. A video summarizing the methodology and the validation in both simulation and real hardware is submitted along with this paper.

Index terms

Reinforcement Learning Machine Learning for Robot Control

Related papers