Efficient and Compliant Control Framework for Versatile Human-Humanoid Collaborative Transportation
Embedded Files
In this work we developed an MPC-QP control framework for heavy co-transportation tasks between humans and humanoids. Read more here
Heavy load lifting (16kg)
Compliant response leading to higher efficiency
Passively follows human's lead
Admittance model and Novel Interaction Linear Inverted Pendulum (I-LIP) model for footstep planning using MPC for compliance and stability
Object informed low-level QP joint torque control and stiffness modulation for heavy lifting (16kg) and tracking of desired distance between the humanoid and the object
C++ ROS2 framework with shared memory based deployment on the humanoid
Implementation experience earned:
C++. Python, MuJoCo, ROS2, IP networking, Optimization (Casadi, qpOASES), F/T sensors, Optitrack motion capture, etc.
Bipedal Navigation in Cluttered Spaces
In this work we developed an MPC-QP control framework for safe navigation in cluttered spaces. This framework provides closed loop recursive feasibility guarantees for the complete framework.
N-step caturability based recursively feasible MPC for navigation in constrained environments.
Switching.mp4Capturability-aware recursively feasible sequential MPC for safe navigation in multiple intersecting safe corridors.
Closed loop recursive feasibility of the MPC-QP based whole body controller using control barrier functions.
Turning.mp4A closed loop recursively feasible MPC-QP framework for safe navigation in cluttered spaces. Includes turning manuevers and corresponding recursive feasibility guarantees.
Development and Control of Leg Exoskeleton
We designed and developed a 7Dof lower limb exoskeleton for gait rehabilitation of stroke patients
Human gait tracking
Adaptive assist-as needed controller for gait rehabilitation using the developed exoskeleton. The controller creates a force field around the leg to provide assistance as needed. Read here
Contributions:
Designed and manufactured a lower limb exoskeleton for gait rehabilitation. Applied PD control using the dSPACE MicroLabBox.
Designed an adaptive assist as needed controller using Lyapunov theory to provide gait assistance without the knowledge of inertial parameters of the human leg
Estimated human joint torques using exoskeleton pose and F/T measurements.
Motion Planning of Handspring Maneuver
Developed a constant torque strategy for infinite handspring maneuvers of an acrobat robot
Derived an underactuated dynamic model of an Acrobat performing front handsprings, validated it, and performed motion planning to obtain near-periodic trajectories using only constant-torque actuation during the cycle
Developed a robust torque perturbation strategy enabling infinite continuous handsprings and a 20±5% error margin in the initial configuration of the acrobat.
Page updated
Google Sites
Report abuse