Dr. Zhichao Liu

I'm a Postdoctoral Research Scientist at The Institute for Integrative & Innovative Research (I3R), University of Arkansas in Fayettiville, AR. I work with Prof. Mehran Armand and Prof. Alejandro Martin-Gomez on emboided medical robots and imaging systems. Previously, I worked as a Postdoc in University of California Riverside in Riverside CA with Prof. Konstantinos Karydis , I lead research efforts in robotic learning, perception, and embodied AI navigation, with a focus on applications in precision farming and field autonomy. I obtained my PhD in Electrical and Computer Engineering at University of California Riverside in 2023.

Email  /  Scholar  /  LinkedIn

Research

I am interested in high-level robotic autonomy across various locomotion modalities, leveraging AI-driven perception and path planning in natuaral challenging environments, while enabling human-level dexterous manipulation in complex physical interactions.

This paper addresses the challenge of developing a multi-arm quadrupedal robot capable of efficiently harvesting fruit in complex, natural environments. We introduce the first three-arm quadrupedal robot LocoHarv-3 and propose a novel hierarchical tri-manual planning approach, enabling automated fruit harvesting with collision-free trajectories. Our comprehensive semi-autonomous framework integrates teleoperation, supported by LiDAR-based odometry and mapping, with learning-based visual perception for accurate fruit detection and pose estimation.

This paper presents an embodied AI system designed for human-in-the-loop navigation using a wheeled mobile robot. The authors propose an effective method to monitor the robot’s current plan, detect significant environmental changes affecting its trajectory, and query a human for feedback. The system translates human feedback provided in natural language into local navigation waypoints, integrating these into a global planning framework that uses a semantic feature map and obstacle map.

This paper explores the design and implementation of a bimanual UAV for autonomous avocado harvesting, aimed at improving efficiency in precision agriculture. Ground-based robots are often ineffective for crops like avocados due to unstructured terrain and hard-to-reach fruits. To address this, the UAV is equipped with two arms: a gripper and a fixer arm. The gripper detaches avocados by applying rotational motion, while the fixer arm stabilizes the peduncle, which can store elastic energy, preventing detachment. The UAV employs visual perception and learning to detect avocados and estimate their pose, which informs the bimanual manipulation planner.

The article presents an impact-resilient aerial robot (s-ARQ) equipped with a compliant arm designed to reduce collision impact and sense contacts. The robot includes a real-time contact force estimator and a non-linear motion controller that enables it to handle collisions during aggressive maneuvers and stabilize after high-speed impacts with walls. Additionally, a new collision-inclusive planning method, which prioritizes contacts to assist navigation in cluttered environments, is introduced.

This study explores the impact-resilient capabilities of MAVs with passive springs in their compliant arms, using dynamic modeling to show that compliance improves impact resilience. Extensive tests demonstrate the MAV's ability to stabilize after high-speed, large-angle wall collisions. The study also includes comparisons with rigid MAVs to assess the trade-offs of adding compliance to the robot's frame.

This paper develops and validates a closed-loop controller for soft pneumatic actuators in a pediatric wearable robotic device, achieving accurate trajectory tracking for two-degree-of-freedom shoulder motion in an infant-sized mannequin.

The paper provides a review of recent advances in the application of Koopman operator theory for the modeling and control of soft robots.

This work presents the development of a Soft Aerial Gripper (SoAG) equipped with a horizontal soft gripper and pneumatic system, designed to safely catch aerial micro-robots mid-air by mitigating aerodynamic disturbances, with experimental results demonstrating its effectiveness in both static and dynamic grasping scenarios.

This article introduces a hierarchical framework for collision-inclusive motion planning and control for impact-resilient mobile robots navigating in unknown and cluttered spaces. The framework integrates a local collision recovery and trajectory replanning strategy with a global search-based planning algorithm that exploits potential collisions to improve performance metrics like energy efficiency and computational time, showing efficacy through extensive simulation and experimental testing.

This work presents a closed-loop control system for a soft wearable device with pneumatic actuators, using proprioceptive feedback from IMUs to assist infant reaching tasks by regulating joint motion based on anthropometric data and desired trajectories, with successful experimental results demonstrated on a mannequin.

This paper introduces a collision-inclusive motion planning approach for impact-resilient mobile robots, featuring a deformation recovery controller and post-impact trajectory replanner that optimize recovery after collisions and generate efficient trajectories, with experimental validation on an omnidirectional robot equipped with Hall effect sensors.

This paper presents a closed-loop trajectory tracking control scheme for soft pneumatic legged robots, demonstrating precise control of body height and orientation over flat terrain using a compact pneumatic regulation system, with experiments showing successful tracking of straight, curved, and variable-height trajectories, laying the foundation for autonomous navigation.

This paper introduces a collision-resilient quadrotor with a compliant arm design that absorbs shocks, alongside a novel collision detection method using Hall sensors and a recovery control strategy for smooth trajectory tracking post-collision, demonstrating the robot's ability to detect and recover from high-speed impacts with various obstacles, including unstructured surfaces and moving objects.

This paper presents a novel sampling-based online planning algorithm for navigating unknown, obstacle-cluttered environments by allowing potential collisions to be harnessed rather than avoided, featuring a joint optimization function that evaluates collision effects and a state expansion pruning technique to reduce search space, with experimental validation on a collision-tolerant holonomic wheeled robot exploring trade-offs between risk levels, collision decision-making, and trajectory efficiency.

This paper introduces SoRX, a novel soft robotic hexapedal robot utilizing a new 2-degree-of-freedom soft pneumatic actuator, achieving forward speeds of up to 0.44 body lengths per second while successfully navigating various challenging terrains using an alternating tripod gait.

This project develops and evaluates the first actuated wearable device for infants with or at risk for motor impairments, featuring four pneumatic actuators to actively assist upper extremity movement, with positive results on performance, wearability, and safety during vertical reaching tasks.