top of page


Seqeuntial Trajectory Optimization for Externally-Actuated Modular Manipulators with Joint Locking

We introduce a novel trajectory planning method for externally-actuated modular manipulators (EAMMs). Joint-locking feature allows effective balancing of the payload capacity and dexterity of the robot but significantly complicates the planning problem by introducing binary decision variables. To address this challenge, we leverage the problem's intrinsic structure, i.e., the payload at the end-effector being enhanced by merely locking its immediate connected links; this allows us to break down the complex planning problem into a series of manageable subproblems and solve them sequentially. We demonstrate the effectiveness of this approach through various simulations and experiments. (ICRA 2024, Best Unmanned Aerial Vehicles Paper Award - Finalist) 

LASDRA External Wrench Estimation

We introduce an innovative external wrench estimation method tailored for the LASDRA system equipped with a joint locking mechanism. Traditional estimation methods face limitations due to unknown joint locking torques. Additionally, when joints are locked, the system's degrees of freedom are reduced, leading to kinematic degeneracy which renders Jacobian-based wrench estimation methods ineffective. To overcome these challenges, we have developed both distributed and centralized external wrench estimation approaches that effectively address the issue of unknown joint locking torques. Our 4-link LASDRA system demonstrates the capability to estimate external forces with a root mean square error (RMSE) of approximately 1.5N. (IROS2023)

LASDRA with Joint Locking 

We present experimental results of LASDRA with a joint locking mechanism to verify the performance and effectiveness of the system with trajectory tracking and object transportation experiments. In the LASDRA system, the payload of the end-effector is determined by the payload of the distal link only, thus we adopt a lightweight joint locking mechanism to lock adjacent links so that it can increase the maximum payload by distributing the payload to the neighborhood, at the same time, to meet a tight weight budget. 3-link LASDRA system can transport a 1.28 kg object with joint locking while the maximum payload without joint locking is 0.4 kg. (IEEE Spectrum (2022))

RVM (Robot-based Vibration Suppression Modules) 

RVM is an easily snap-attachable/detachable two-rotor module with all sensors (i.e., IMU), battery and computing onboard, so that, by simply attaching many of them distributedly on a very slender flexible object, they can help to transport/manipulate the object while providing load-sharing and vibration-suppression capabilities. To maximize these capabilities under the current battery-rotor technology limitation, RVM two-rotor design is optimized. Further, how to optimally place multiple RVMs on the bar-like flexible object with sensing/actuation uncertainties taken into account is also elucidated. (ICRA2020, RSS Pioneer workshop) 

Flying-LASDRA: Autonomous Outdoor Flying  

​We perform outdoor autonomous flying experiment of flying-LASDRA (aka f-LASDRA) system, constructed with multiple ODAR-8 links ( connected via cable with each other. Each ODAR-8 can compensate for its own weight, rendering f-LASDRA scalable. Utilizing SCKF with IMU/GNSS-module on each link and inter-link kinematic-constraints, we attain estimation accuracy suitable for the stable control of flying while maintaining internal shape by using only standard GNSS, not RTK-GPS (<5cm: cf. 1-5m w/ GNSS).  (ICRA2019, IEEE Spectrum (2019)) 

LASDRA (Large-Size Aerial Skeleton w/ Distributed Rotor Actuation) 

We present experimental results of a novel robotic system, LASDRA (large-size aerial skeleton w/ distributed rotor actuation), which consists of two 1.5m fully-actuated links, making 3m long 6-DOF manipulation system with weight of only 10kg, an order of magnitude lighter than that of similar-size electrical/hydraulic motor robots.  The LASDRA system is scalable, can be made arbitrarily long, dexterously-articulated, yet, still light/slender .  Shown here are its  end-effector trajectory tracking and valve turning experiments. (ICRA2018)

ODAR (Omni-Directional Aerial Robot)

The ODAR (omni-directional aerial robot) is a fully-actuated aerial platform, that can generate arbitrary control wrench in se(3) by six opportunistically distributed rotors, each driven by reversible ESC (electronic speed controller) so that it can realize such powerful behaviors impossible with conventional multi-rotor flying platforms as resisting sideway gust while keeping its attitude and exerting downward pushing force larger than its own weight. (IROS2016, TMECH2018, also featured in IEEE Spectrum (2017, 2018))

SmQ (Spherically-Connected multi-Quadrotor) Platform

SmQ platform is a new platform for aerial manipulation, which consists of a tool (or frame) with multiple quadrotors connected to that by spherical joints, acting as distributed rotating thrust generators.  Lyapunov control design is performed while taking into account the spherical joint limits in a form of real-time constrained optimization. (IROS2015, TRO2018, also featured in IEEE Spectrum (2015, 2017))

QM (Quadrotor-Manipulator) System

We show that, similar to rigid-body dynamics in SE(3), the QM-system also consists of two decoupled dynamics: 1) center-of-mass dynamics in E(3) with under-actuation and gravity; and 2) the “internal rotational” dynamics, which assumes the form of standard manipulator dynamics with full-actuation and no gravity. Utilizing this,  we design backstepping control, that allows us to achieve coarse(platform)-fine(manipulator) control as well as cooperative manipulation of multiple QM systems. (ICRA2014, ICRA2015)

QT (Quadrotor-Tool) System

We propose a novel control framework to enable a quadrotor to operate a tool attached on it.  We fully characterize the internal dynamics of the spatial quadrotor tool operation, which arises due to the quadrotor’s under-actuation, and elucidate a seemingly counter-intuitive necessary condition for the internal stability, that is, the tool-tip should be located above the quadrotor’s center-of-mass. We further manifest that this internal dynamics can exhibit finite-time escape and propose a stabilizing action to prevent that.  (DSCC2012, IROS2013, Automatica2015)

bottom of page