US-based Sunday Robotics has unveiled its new AI model, ACT-2, enabling its Memo home robot to fold laundry with unprecedented accuracy. The system achieved a 99.1% success rate in real-world homes it had never encountered, handling unfamiliar clothing items. This breakthrough represents a significant leap forward in developing autonomous robots capable of operating effectively outside controlled laboratory settings.
Overcoming a Fundamental Robotic Challenge
Handling deformable objects like clothing has long been a difficult task in robotics due to their lack of a fixed structure. Most previous systems required training in specific environments or on pre-defined types of garments. Sunday Robotics' innovation directly addresses this challenge, aiming to create a more versatile and adaptable home assistant.
The Power of Generalization in AI
The ACT-2 model operates on the principle of “generalization,” allowing it to apply learned skills to new situations without retraining. This enables the Memo robot to adapt to different homes, surfaces, and clothing items it has never seen before. This adaptability is the key differentiator between this new generation of robots and their more rigid predecessors.
The robot's initial training involves human demonstrators wearing sensor-equipped gloves that transmit hand movements to the system. Sunday Robotics then uses an internal fleet of robots to continuously practice and refine these skills. This iterative learning process is crucial for building the reliability needed for deployment in consumer homes.
Rigorous Real-World Testing and Performance
To validate its system, the company conducted extensive tests inside various real homes rather than simulated environments. The Memo robot was tasked with handling nine different clothing types under diverse conditions, including various folding locations and surface colors. This rigorous methodology was designed to prove the robot's performance in unpredictable, everyday settings.
The results demonstrated a high degree of competence, with the robot successfully completing 778 out of 785 folding attempts. This translates to an impressive 99.1% success rate, which the company presents as strong evidence of the system's readiness. Such a high level of reliability is a critical milestone for the home robotics industry.
Setting a New Industry Standard
Alongside its technological announcement, Sunday Robotics proposed a new benchmark for evaluating home robot performance called Solve. This standard aims to unify testing protocols by specifying conditions, the number of attempts, and success rate metrics. The goal is to foster greater transparency and allow for more accurate comparisons between different robotic systems.
Future Roadmap and Competitive Landscape
The company plans to launch a pilot program for the Memo robot in the fall of 2026, deploying it in select homes for daily use. Looking ahead, Sunday Robotics intends to expand Memo's capabilities to include floor cleaning, coffee preparation, and general tidying. This roadmap illustrates a vision for a multi-functional home assistant that evolves over time.
This announcement places Sunday Robotics in a competitive field, with players like Tesla, 1X, and Weave Robotics also developing AI-powered home robots. The race to create autonomous assistants for daily chores is accelerating as AI and robotics technologies advance. The market for such devices is expected to experience significant growth in the coming years.
Sunday Robotics' achievement with the ACT-2 model signals a pivotal shift in the home robotics industry, moving the focus from isolated skills to reliable performance. By achieving high accuracy in unpredictable real-world environments, the company is helping to set a new benchmark for success. This progress could be a crucial step toward the widespread adoption of truly autonomous and helpful robots in our homes.