Toyota Research Institute Has Released A New Product! Soft Humanoid Robot Is Here!

Humanoid robots have been thrust into the spotlight over the past year, with companies lining up to release their own humanoid products. Most of them have a typical humanoid appearance, using arms and claws to handle objects, and hard legs as their way of walking.

But recently, Japan's Toyota Research Institute (TRI) launched a new robot, Punyo, and expressed its hope that Punyo will push humanoid robots forward.

Punyo is innovative in the design concept and operation methods of robots. It has no legs, and so far, the TRI team has been working on the robot's torso and developing manipulation skills.

Design concept: serving human daily life

Traditional industrial robots are mostly used in workshop operations, assembly and other tasks to improve production efficiency and reduce labor intensity. In the future, service robots may enter more homes, directly facing and serving the daily needs of ordinary people.

TRI researchers said Punyo's goal is to become a robot that "helps people complete daily tasks at home and elsewhere."

This design concept determines that Punyo needs to be flexible, soft, and safe. Because in order to enter the complex and changeable home environment, there cannot be a hard and rigid mechanical arm like a traditional industrial robot. Otherwise, it will give people a sense of danger and make it impossible to complete various daily items operating tasks. This is somewhat similar to the design idea of SoftBank's robot Pepper, which focuses on how to make robots more integrated into human life.

Service-oriented applications also require Punyo to learn a variety of daily skills, not just perform a single operation on the factory assembly line. This requires giving robots strong learning capabilities and mastering the operation methods of various daily tasks by observing and imitating human demonstrations.

For humanoid robots, manipulation using the entire body is tricky because balance is a challenge. However, TRI researchers designed its robot to do just that.

"Punyo does things differently. Using its entire body, it can carry much more than simply pressing with an outstretched hand," added Andrew Beaulieu, one of TRI's technical leads for full-body manipulation. "Softness, tactile sensitivity and the ability to make a lot of contact facilitate better manipulation of objects."

Soft and hard body

To achieve a flexible and soft robot design, TRI adopted a mechanical arm design that combines hard and soft. Punyo's hands, arms, and chest are covered in compliant materials and tactile sensors that both sense outside contact, and the soft materials allow the robot's body to conform to the objects it's manipulating.

This is a typical design idea for many current soft robots.

At the same time, under the soft shell, Punyo also retains two "hard" mechanical arms as skeletal supports, as well as a torso frame and waist actuator to provide mechanical support and precise control. This combination of hard and soft design combines the mechanical advantages of traditional robots with the soft characteristics of soft robots.

Specifically, the airbags on Punyo's arms can adjust the internal pressure to become harder or softer as needed. While ensuring a certain mechanical rigidity, it also provides about 5 cm of compliance. The "claw" also uses a high-friction latex airbag design. The camera in the palm of the hand can sense the size of the external force by observing the surface deformation of the airbag. The entire arm can be bent and rotated, and the air bags are connected to each other, which allows the force to be transmitted smoothly and prevents the robot from "breaking the arm".

Strong learning ability

To adapt to the changing tasks in the home environment, Punyo must have strong learning abilities.

According to the TRI team, Punyo learned a contact-rich policy using two methods: a diffusion strategy and example-guided reinforcement learning. TRI announced its approach to proliferation policy last year. With this approach, robots use human demonstrations to learn robust sensorimotor strategies for difficult-to-model tasks.

Example-guided reinforcement learning is an approach that requires modeling a task in a simulation and guiding the robot's exploration through a small set of demonstrations. TRI says it uses this learning to implement robust operating strategies for tasks that can be modeled in simulations.

When a robot can see these tasks demonstrated, it can learn them more efficiently. It also gives the TRI team more room to influence the style of movement the robot uses to complete its tasks.

The team used adversarial motion priors (AMPs), traditionally used to stylize computer animated characters, to incorporate human motion mimicry into their reinforcement pipeline.

Reinforcement learning does require teams to model tasks in simulations for training. To do this, TRI uses a model-based planner for demonstrations rather than remote operations. It calls this process "plan-guided reinforcement learning."

TRI claims that using the planner can make long-distance missions that are difficult to operate remotely possible. The team can also automatically generate any number of demos, reducing its pipeline's reliance on human input, which brings TRI closer to increasing the number of tasks Punyo can handle.

Although the Punyo software service robot is still in its infancy and its performance in all aspects needs to be improved, its application prospects are broad, and Punyo's design concept and technical route also provide new ideas for the industry.