Yang(Kevin) Liu

I'm a senior undergraduate student at the UMN, pursuing a B.A. in Computer Science. Currently, I am fortunate to be advised by Prof. Heng Yang at the Harvard Computational Robotics Group.

Previously

I was extremely fortunate to be advised by Hanchen Cui and Karthik Desingh.
I was a research intern at Shanghai Qi Zhi Institute, where I had the privilege of being advised by Prof. Hang Zhao and Ziwen Zhuang.
I am also lucky to work closely with Liang Wang.

Email / Scholar / Twitter / Github

Research

I'm interested in robot learning, focusing on perceptive locomotion, humanoid loco-manipulation, manipulation, and visuo-tactile learning.

	PILOT: Guiding Long-Horizon World-Model Planning for Robot Manipulation with Visual Subgoals and Policy Priors Yang Liu^, Hanchen Cui^, Karthik Desingh ^* Equal contribution Under review project page / arXiv PILOT guides long-horizon robot manipulation with video-predicted visual subgoals, policy-initialized local optimization, and an action-conditioned world model.
	Trajectory Optimization via Schrödinger Bridge Sampling Mattia Mosso^, Yang Liu^, Heng Yang ^* Equal contribution Under review arXiv A diffusion-sampling framework for trajectory optimization with hard path and terminal constraints, demonstrated on contact-rich locomotion and manipulation.
	PUMA: Perception-driven Unified Foothold Prior for Mobility Augmented Quadruped Parkour Liang Wang, Kanzhong Yao, Yang Liu, Weikan Qin, Jun Wu, Zhe Sun, Qiuguo Zhu RA-L 2026 project page / arXiv PUMA enables robust quadruped parkour by learning a perception-driven foothold prior from depth and proprioception.
	Bridging Perception and Planning: Towards End-to-End Planning for Signal Temporal Logic Tasks Bowen Ye, Junyue Huang, Yang Liu, Xiaozhen Qiao, Xiang Yin, ICRA 2026 arXiv An end-to-end perception-to-planning model that maps raw images to feasible trajectories satisfying formal STL constraints.

Projects

Here are some fun projects I've worked on outside robotics.

Feed-forward 3D Reconstruction for Spatial Semantic Understanding
Fall 2025
poster / code

Semantic-aware Feature-3DGS compression enables efficient scene understanding by pruning Gaussians based on their semantic-loss contribution while stabilizing features via local kNN consistency.

Miscellanea

Teaching	Teaching Assistant, CSCI 1933 Fall 2025
Service	Reviewer, International Conference on Robotics and Automation(ICRA) 2026
Hobby	⚽ Soccer; Go(Weiqi) ⚫️⚪️

Design and source code adapted from Leonid Keselman's Jekyll fork of Jon Barron's website.

Research

Projects

Miscellanea

Teaching

Service

Hobby