Intelligent Mechatronics: Robot Manipulation

We study robot manipulation through data-efficient learning, tactile interaction control, and actuation design. Our emerging program connects simulation and physical experiments to develop robots that learn efficiently and interact reliably.

Concept illustration of demonstrations, digital twins, and synthetic data generation.

Data-Efficient Learning

Digital twins and mixed real and simulated data support efficient adaptation to new tasks.

Concept illustration of tactile exploration, contact sensing, and force control.

Tactile Interaction Control

Tactile sensing and force feedback guide stable contact during robotic manipulation.

Concept illustration of twisted-tendon actuation, dexterous hands, and dynamic contact.

Actuation and Dynamics

Actuator design and dynamics studies support controlled motion during robotic manipulation.

Scene Reconstruction

Objects are parsed from a single scene image and reconstructed alongside the static scene.

DEXTERA diagram showing scene alignment, robot-camera calibration, and object scale estimation.

Metric Alignment

Robot and scene calibration aligns object scale and pose for physical deployment.

DEXTERA diagram showing simulator reconstruction, reusable task primitives, and multimodal observations.

Task Construction

Reusable simulation tasks and multimodal observations support policy learning.

Scene Observations

Simulated and real workspaces are compared through RGB images, instance masks, and depth.

Comparison of reconstructed coconut-water carton, thermometer, and toy bus across methods.

Object Reconstruction

Reconstruction methods are compared using the shape and appearance of three objects.

DEXTERA simulated and real sequences of basket lifting and bowl handover.

Cross-Domain Replay

Simulated and real sequences show basket lifting and bowl handover across interaction stages.

KUKA LBR iiwa 7 robot with a LEAP Hand, cameras, and task board.

KUKA + LEAP Hand

A KUKA LBR iiwa 7 with a LEAP Hand uses cameras and a task board for dexterous manipulation.

Dual-arm OpenArm platform with BrainCo Revo1 hands, RealSense cameras, and task board.

OpenArm + BrainCo Revo1

OpenArm combines two arms and BrainCo Revo1 hands for coordinated bimanual manipulation.

Flexible tactile sensor array attached to a soft gripper with Arduino Nano and readout electronics.

Flexible Tactile Sensing

A soft gripper integrates a flexible tactile array and Arduino Nano electronics for readout.

Robot-mounted tactile gripper beside a laptop displaying its sensor readout.

Gripper Testing

A robot-mounted gripper performs contact experiments and displays live tactile measurements.

Tactile Finger Gripper

Twisted-tendon fingers grasp an object and provide a test platform for tactile manipulation.

Twisted-tendon actuator test bench with electronics and a laptop displaying measurements.

Actuator Test Setup

A benchtop setup combines twisted-tendon actuation and electronics for actuator characterization.

DEXTERA: Real-to-Sim-to-Real Manipulation

Explore the interactive demonstrations and full project details below, or open the DEXTERA project website.

MINIMAX Lab

Intelligent Mechatronics: Robot Manipulation

DEXTERA: Real-to-Sim-to-Real Manipulation