Videos

Pick-and-Place by Octo

torobooctoimpedance control

the robot's behavioral policy was acquired through imitation learning. We collected about 50 samples for each instruction through teleoperation. We used Octo, an open-source generalist robot policy, as a pre-trained model. The pre-trained Octo model was fine-tuned to our collected examples.The learned policy monitors the task progress and generates the next action based on the camera image. Thus, it can complete the task even if it fails or is disturbed during execution.

Wiping table with Deep Predictive Learning

torobodeep predictive learningimpedance control

This demo showcases our robot performing autonomous table wiping powered by Deep Predictive Learning developed by Ogata Lab at Waseda University. Through several dozen human teleoperation demonstrations, the robot has learned natural wiping motions. The top-left window displays predicted images from the trained model, where blue dots represent the robot’s current focal points, and red dots indicate their projected future positions. With impedance control, the robot maintains an optimal contact force with the table surface, enabling smooth and effective cleaning motions.

Pitching a Ball

toroboposition control

This is a pitching demonstration of new Torobo. New Torobo is lighter than the older version, enabling faster motion such as throwing a ball. The new model will be available in Japan in March 2025 and overseas from October 2025 onward. For more information, please refer to the brochure and the GitHub repositories.

New Torobo and New Torobo Hand in Isaac Sim and MuJoCo

isaac simmujoco

This is a demonstration of the new Torobo and new Torobo Hand in Isaac Sim and MuJoCo. To accelerate robotics development, we've open-sourced our simulation models of the new Torobo and new Torobo Hand. These models can be used in both Isaac Sim and MuJoCo. If you're interested in using these models yourself, please visit our GitHub repositories and our blog (available in Japanese only) for more information and setup instructions.

Bipedal Walking by reinforcement learning in Isaac Sim and Validating Trained Policy in MuJoCo

toroboreinforcement learningisaac simmujoco

This is a reinforcement learning process of bipedal walking in Isaac Sim and validation of the trained policy in MuJoCo. Isaac Sim enables massively parallel reinforcement learning and completes training of bipedal walking in a short time. To validate the trained policy in different physical simulations, the policy is transferred from Isaac Sim to MuJoCo.

Hammering a nail

toroboimpedance control

This is a demo of hammering a nail. The instantaneous rebound force from the hammer is absorbed through a combination of the elasticity of the rubber material securing the hammer, the deflection in torque sensors and harmonic gears, backdrivability, and impedance control. This allows the nail to be driven with a certain amount of force.

Sawing wood using impedance control

toroboimpedance control

This is a demo of cutting wood with a saw. When using position control for this task, precise measurement of the cutting amount is necessary. However, by using impedance control, this requirement is eliminated, allowing for successful cutting with only rough commands.

Box transportation using a puppet-type teleoperation device

toroboteleoperationpuppeteering

This is a demonstration of transporting a box using a puppet-type teleoperation device. The device can be operated intuitively as the joint configuration of the robot and the puppet are the same. The mobile base is controlled by a controller attached to the puppet's hand.

Torque and Impedance control of Torobo Hand

torobo handimpedance control

Our Torobo Hand can grasp objects of any shape using torque and impedance control, without prior knowledge of their exact dimensions.

Table wiping using impedance control

toroboimpedance control

The robot can press down on the towel with appropriate force, making it highly compatible with control by generative AI.

Shock absorption using impedance control

toroboimpedance control

Torobo carrying a 25 kg (55 lbs) bag. Its backdrivability and impedance control help absorb a sudden change in external force.

Torobo Hand

torobo hand

This is a 10-axis multi-finger hand using the world's smallest cycloidal reducer developed by Tokyo Robotics. Utilizing an efficient cycloidal reducer enables impedance control. This allows not only to conform to various object shapes but also to prevent jamming fingers.

Whole-body humanoid robot - Torobo

torobo

This video shows the basic performance of humanoid robot “Torobo”, which is used as a research platform for JST’s Moonshot R&D program (Goal 3: Smart Robot that is Close to One Person for a Lifetime).

Shoebox picking with Torobo GTC

logisticsgtc

This is a demo of shoebox picking using Torobo GTC. The system does not require picking stations, which reduces labor costs and enables 24-hour unmanned operation. Additionally, it offers storage efficiency due to its capability of direct picking from shelves approximately 3.9 meters high.

Shoebox picking with Torobo GTC (with depth)

logisticsgtc

This is a demo of shoe box picking using Torobo GTC. It can also handle boxes where there are multiple layers in the depth direction (for example, automating operations in fixed locations previously performed manually).

Server check demo at datacenter

tolondatacenter

This is a demo of a remote inspection solution in a NTT Communications' datacenter. The robot can move autonomously, opening and closing the aisle cap, remotely control the 6-axis posture of the arm, and take pictures inside the server rack using the end-of-arm camera.

Autonomous Case-handling Robots

tolonlogistics

This is a demo of automatic case handling using a mobile manipulator "Tolon" and a Mobile Gripper(TM). They recognize boxes and conveyors, and conduct the task autonomously

Whole-body humanoid robot - Toala

toala

This video shows the basic performance of a humanoid robot Toala. By estimating the external force from the motor current value, impedance control is possible without joint torque sensors.

Mobile Manipulator - Tolon

tolon

Tolon is a mobile manipulator that can work in narrow spaces thanks to its asymmetrical arm arrangement and omni-directional mobile base. It is ideal for the development of autonomous robots that automate material handling, short-distance transport, security, patrolling, etc.

Remote force control using Torobo

toroboteleoperationbilateral control

This is an application of a bilateral control system, which can transmit external force of a humanoid robot to its operator. This system is a result of joint development since we were selected for the Open Innovation Program by NTT Communications in 2019.

Remote Control of Humanoid Robot Toala

toalateleoperationmotion capture

This is a demo of remotely controlled Toala, a humanoid robot capable of force control without joint torque sensors. Noitom's Perception Neuron PRO is used as the control device. This system is provided as an optional extra of Toala.

Palletizing with Mobile Gripper

mobile gripperlogistics

This is a demo of automatic palletizing with a Mobile Gripper(TM). By using left and right independent movable grippers and force detection, cardboard boxes are stacked without gaps.

Bilateral control demo of a remote humanoid

flagshipteleoperationbilateral control

This is a demo of force and position control of a humanoid robot from a remote location using bilateral control over a network. This is the result of joint work with NTT Communications.

All our videos can be found on Youtube. Please check them out.

See all videos