I am honored to receive Viterbi Fellowship and AME Department Fellowship from USC. I am fortunate to collaborate with Prof. Nikolay Atanasov (UC San Diego), and Dr. Guillaume Bellegarda (EPFL, Switzerland)
My research interests span optimization, control, and learning approaches for dynamic robotics including trajectory optimization, deep learning, and real-time optimization-based control.
Publications
Projects for Optimization and Control
The first ever double barrel roll achieved by A1 robot
Performing highly agile acrobatic motions with a long flight phase requires perfect timing, high accuracy, and coordination of the full-body motion. To address these challenges, we present a novel approach on timings and trajectory optimization framework for legged robots performing aggressive 3D jumping. In our method, we firstly utilize an effective optimization framework using simplified rigid body dynamics to solve for contact timings and a reference trajectory of the robot body. The solution of this module is then used to formulate a full-body trajectory optimization based on the full nonlinear dynamics of the robot. This combination allows us to effectively optimize for contact timings while ensuring that the jumping trajectory can be effectively realized in the robot hardware. We first validate the efficiency of the proposed framework on the A1 robot model for various 3D jumping tasks such as double-backflips off the high altitude of 2m. Experimental validation was then successfully conducted for various aggressive 3D jumping motions such as diagonal jumps, barrel roll, and double barrel roll from a box of heights 0.4m and 0.9m, respectively.
Performing highly agile dynamic motions, such as
jumping or running on uneven stepping stones has remained
a challenging problem in legged robot locomotion. This paper
presents a framework that combines trajectory optimization
and model predictive control to perform robust and consecutive
jumping on stepping stones. In our approach, we first utilize
trajectory optimization based on full-nonlinear dynamics of
the robot to generate periodic jumping trajectories for various
jumping distances. A jumping controller based on a model
predictive control is then designed for realizing smooth jumping
transitions, enabling the robot to achieve continuous jumps on
stepping stones. Thanks to the incorporation of MPC as a
real-time feedback controller, the proposed framework is also
validated to be robust to uneven platforms with unknown height
perturbations and model uncertainty on the robot dynamics.
Projects for Robot Learning
Robustness to environment noise of foot disturbances
In this paper, we consider a general task of
jumping varying distances and heights for a quadrupedal robot
in noisy environments, such as off of uneven terrain and with
variable robot dynamics parameters. To accurately jump in such
conditions, we propose a framework using deep reinforcement
learning that leverages and augments the complex solution of
nonlinear trajectory optimization for quadrupedal jumping.
While the standalone optimization limits jumping to take-off from
flat ground and requires accurate assumptions of robot dynamics,
our proposed approach improves the robustness to allow jumping
off of significantly uneven terrain with variable robot dynamical
parameters and environmental conditions. Compared with
walking and running, the realization of aggressive jumping on
hardware necessitates accounting for the motors’ torque-speed
relationship as well as the robot’s total power limits. By
incorporating these constraints into our learning framework, we
successfully deploy our policy sim-to-real without further tuning,
fully exploiting the available onboard power supply and motors.
We demonstrate robustness to environment noise of foot disturbances of up to 6 cm in height, or 33% of the robot’s nominal
standing height, while jumping 2x the body length in distance.
The realization of highly dynamic jumping on legged robot to particular targets at high accuracy is a challenging task.
This inaccurate transfer normally comes from uncertainties and unknown dynamics involved in contact dynamics and
hardware model. In this paper, we are inspired by nature of practice make perfect on human and animal jumping to
propose a framework based on iterative learning control to address this challenging problem. Our approach allows the
robot to apply what it learns from a simple jumping task to accomplish multiple complex tasks within a few trails directly
in hardware. In addition, the iterative learning can be efficiently formulated as Quadratic Programs (QP), enabling fast
solving time of less than 1 second for each trail. We validate the method via extensive experiments in A1 model and
hardware for various jumping tasks. From a simple short jump (e.g. forward 40cm), our learning approach enables
robot to achieve multiple challenging targets such as jumping on high box of 20cm, jumping farther to 60cm, as well
as jumping while carrying an unknown mass of 2kg. Our method enables the robot to reach desired position and
orientation targets with the approximate errors of 1cm and 1 degree, respectively.