Embedded Files
My research focuses on the learning-based modeling and control with application to robotics. In particular, I focus on the underactuated robots, which possess fewer control inputs than the degrees of freedoms. My research interests include:
Robotics and autonomous systems (autonomous vehicles, manipulation, robot dynamics and control)
Dynamic systems and control (machine learning-based modeling and control, nonlinear control system design, design and control of mechatronic systems)
Learning-based sensor fusion and posture analysis (learning-based sensor data analysis and modeling, gait/posture detection and prediction)
For the uncertainties and modelling errors, a Gaussian Process based machine learning framework has been proposed to deal with modelling and control. The learning based method is integrated with EIC-based control for underactuated balance robots. Simultaneously we can achieve trajectory tracking and balance control.
Quanser Rotary Inverted Pendulum
A ski-stunt maneuver is a type of aggressive vehicle motions in which a four-wheel vehicle runs on two wheels on one side, and the other two wheels are lifted in the air. It is a challenging task even for skilled car drivers to perform a ski-stunt maneuver. We present the safety-guaranteed motion control of autonomous ski-stunt maneuvers. The experimental results confirm that the vehicle can successfully initiate the ski-stunt maneuver to safely navigate among obstacles and narrow passes and then switch to normal driving.
(a) Indoor experiment setup. (b) Vehicle runs on a narrow bridge by performing a ski-stunt maneuver. (c) Snapshots of ski-stunt maneuver initiation process by steering. The labels in (c) illustrate the stages in four-wheel to two-wheel transition process.
As a single-track mobile platform, bikebot (i.e., bicycle-based robot) has attractive capability to navigate ,through narrow, off-road terrain with high speed. However, running crossing step-like obstacles creates challenges for intrinsically unstable, underactuated bikebots. This work presents a novel autonomous bikebot control with two assistive legs to navigate crossing obstacles.
Bikebot manipulation has advantages of the single-track robot mobility and manipulation dexterity. We present a coordinated pose control of mobile manipulation with the stationary bikebot. The challenges of the bikebot manipulation include the limited steering balance capability of the unstable bikebot and kinematic redundancy of the manipulator. A coordinated planning and control design is presented to determine the balance-prioritized posture control under kinematic and dynamic constraints. The results confirm the feasibility to use the bikebot manipulation for plant inspection with end-effector position and orientation errors about 5mm and 0.3 degs, respectively.
Video showing the effectiveness of the proposed control scheme.
Video showing the cooridnated manipulation on a stationry baicycle
Using a recurrent neural network (RNN) with long short-term memory (LSTM) cells. to identify the human/animal gait and predict the poses in the real time with only IMU (Inertial Measurement Unit) measurement available.
Page updated
Google Sites
Report abuse