Videos

Tactile sensing, manipulation, and robot learning in motion.

Selected demonstrations from my research and collaborations. Each recording links to a project page with methods, publications, and evaluation details.

Browse quick GIF previews.

Sensing skins and sensor readouts

Six demonstrations from my IROS lightning talk. Each video preserves the physical footage above and the sensor visualization below, with their original timing.

Physical demonstration + sensor visualization

Textile skin: contact avoidance

The robot responds to contact on its textile-covered arm. The lower panel shows the recorded taxel visualization from the same presentation video.

Physical demonstration + sensor visualization

Printed skin: opening a door

Spot uses its printed arm skin during a door-opening task while holding a tray. Physical footage and the recorded sensor visualization appear together.

Physical demonstration + sensor visualization

Textile skin: human interaction

Hand contact with the textile skin is shown above the corresponding color-coded taxel view.

Physical demonstration + sensor visualization

Printed skin during grasping

Spot grasps an object with the printed skin installed on its arm. The recorded taxel view remains visible throughout the clip.

Physical demonstration + sensor visualization

Textile skin: door contact

The textile-covered arm interacts with a door, shown together with its recorded taxel visualization.

Physical demonstration + sensor visualization

Printed skin: table contact

The printed skin contacts a table. The lower panel makes the recorded taxel response visible alongside the physical interaction.

Robots in action

Tactile sensing, bimanual teleoperation, assembly, and human–robot interaction on physical systems.

Teleoperation

SPARK-Remote bimanual manipulation

An operator performs a bimanual bag task using the SPARK-Remote system. This recording shows human teleoperation.

Physical study video

AugInsert assembly study

Real-world augmentation study video from our AugInsert project. The paper also evaluates visual-force policies in simulation.

Exploratory physical run

Astra / GPT-6 on the Thunder UR5e

An accelerated recording of one blue-marker pickup using wrist-camera observations. The original run took about 15 minutes; this is not a reliability benchmark.

Physical demonstration

Natural-language interaction with Spot

A recorded SpotNLP sequence combining natural-language commands, person following, and semantic navigation.

Simulation studies

Recorded experiments in humanoid control and multi-agent coordination. These clips show simulation results and development diagnostics.

Isaac Lab simulation

G1 walking policy

Walking rollout from the frozen simulation evaluation, seed 2042. Physical robot transfer has not been established.

Simulation · assisted

G1 supported pickup

Supported-reach pickup development trace, seed 5146. Assistance is present; unassisted pickup remains unvalidated.

Multi-agent simulation

Alem coalition behavior

Coalition behavior in a deterministic multi-agent replay, seed 10008. The clip illustrates the coordination setting.

Multi-agent simulation

Comparing coordination strategies

Coordination methods shown side by side in the same simulated setting. The held-out study did not establish an efficiency advantage.

Sensor design

A closer look at the geometry behind robot-wide tactile coverage.

CAD visualization

UR5e tactile-skin walkthrough

A 34-second walkthrough of the UR5e skin geometry. Physical fit through motion, fabrication, and sensing performance remain to be tested.