The first place where robots can make money by folding clothes: the laundry room. Practitioners areWeave RoboticsofIsaacs, specially built for housework, is currently on duty at Tumble Laundry, a paid laundry room. It is worth mentioning that Weave Robotic was founded by the former Apple team and has completed three rounds of financing before the product was officially launched.
Although many robots are now learning to fold clothes, most of them are still showing off their skills at the booth, and Isaacs has reached the next level and is the first general-purpose robot that can fold paid clothes.
So what is the workflow of the robot?
Not only can I fold clothes but I can also store them
First of all, the laundry room is responsible for taking on customers' laundry needs and completing the cleaning and drying of clothes. Next, it is Isaacs's work scope. They are responsible for the folding link that is most labor-intensive and requires certain regularity.
Judging from the details of the capabilities of Isaacs itself, its folding service is not extensive. Weave Robotic has clear standards for qualified clothing folding: folded clothing needs to be neat and tidy.
Take shirts as an example. They need to be uniform in shape, without messy corners, and stacked in the same direction. For styles with collars, the collars should be facing upwards.
After folding the clothes, they will also store the clothes and keep the work surface tidy.
If he can meet these work standards, Isaacs will be one step ahead of others as a clothes folder and will be given priority!
From a technical perspective, the stable operation of Isaacs is inseparable from Weave Robotic’s triple technical support:
First of all, it is equipped with a visual-language-action (VLA) model independently trained by the team, which can accurately identify the type of clothing, determine the position of folding corners, and provide a visual and decision-making brain for accurate folding."
Secondly, the team built a high-performance network stack for it. When Isaacs encounters complex situations (such as special clothing materials and messy placement), human operators can intervene through remote assistance to ensure that the task is successfully completed. Currently, its early prototype can achieve 70%end-to-endAutonomous folding, manual assistance is only activated when necessary;
Finally, a complete set of data pipelines gives Isaacs the ability to continuously learn. Relevant data from each completed folding task will be used to train a new generation of VLA model, making it more efficient and accurate when processing different materials and different styles of clothing.
Now, Isaacs collects data in the laundry room, and the robustness of this model will be further improved after training.
In addition, Isaacs is not positioned as a single-function folding robot, but as a general-purpose household robot.
In addition to the current floor-standing clothes folding, in the future it plans to expand more housework capabilities such as clutter sorting and home security to meet the diverse needs of family scenes.
The team also considered details during the design: when Isaacs is idle, its camera will automatically fold and close, and the torso will also be lowered for storage to protect user privacy to the greatest extent.
Taking Isaacs from technology research and development to commercialization is inseparable from the industry accumulation of the two core founders of Weave Robotic.
Former Apple technology executive starts a business making robots
Weave Robotic was co-founded by Evan Winelan and Kaan Dogrusoz, both of whom have experience at Apple.
CEO Evan Winelan graduated from Carnegie Mellon University and served as Apple's chief AI product manager, leading Apple's AI projects and Siri updates.
CTO Kaan Dogrusoz is also an alumnus of Carnegie Mellon University. He once served as Apple’s research director and participated in the development of products such as Apple Watch and iPhone.
After taking up the post at Isaacs, Kaan said:
It's been over a year since we founded Weave. We always have two things in mind:
1. Design a robot that everyone can see at home.
2. Build robots that can get the job done really fast.