On the morning of September 2nd, the "Robot Kindergarten" co-founded by Tashan Technology and the team at Openmind, led by 2024 Turing Award laureate and "Father of Reinforcement Learning" Professor Richard Sutton, officially opened its doors. The first cohort of robots in various forms has been enrolled as "students" at this innovative institution.
The vision behind the Robot Kindergarten is to enable robots to learn continuously, much like children do, through perception, action, feedback, and reflection. Within this environment, robots will interact with the real world and progressively develop their own distinct form of intelligence. Professor Sutton views the kindergarten as a secure learning space where robots can understand their physical structure, comprehend the world around them, and autonomously evolve their cognitive abilities.
Kris De Asis, Senior Researcher at the Openmind Global Research Institute, alongside Hou Guangdong, Vice President of R&D at Tashan Technology, jointly unveiled the first batch of "Robot Kindergarten Collaborative Research Projects." These initiatives address key technical challenges in autonomous robot learning across seven specific directions.
First, hardware designed specifically for learning, focusing on cost-effective, easily maintainable robotic systems that support rapid reset and recovery. Second, tactile and pain perception mechanisms, utilizing touch as the foundational sensory channel for exploration while researching full-body e-skin deployment, pain response systems, and safety protocols.
Third, improved real-time continuous learning algorithms, developing more efficient algorithms capable of running live during operation. Fourth, planning based on continually learned models, investigating how robots can use evolving environmental models for prediction and decision-making in dynamic settings. Fifth, curiosity-driven and directed exploration, examining how internal reward mechanisms can motivate proactive exploration when external feedback is sparse or delayed.
Sixth, sub-problems of feature acquisition, exploring methods for acquiring, representing, and reusing features for general intelligence, including the OAK architecture for defining tasks from an agent's perspective. Seventh, transitioning from design-time to runtime learning, bridging the gap between simulation training and real-world deployment to integrate the complete learning process from development through operation.
The facility is organized into distinct zones tailored to different research needs. The testing area focuses on hardware fundamentals, supporting parallel validation of multiple robot forms with emphasis on long-duration operational robustness. Tashan Technology has already partnered with various ecosystem collaborators to retrofit existing robots with tactile skin technology.
The learning area provides a safe interaction environment adapted for robots, facilitating autonomous exploration through multimodal perception. The interaction zone addresses longer-term objectives, exploring how robots understand and collaborate with other intelligent agents—representing the kindergarten's most ambitious goal of enabling skill acquisition and cooperative abilities.
The Robot Kindergarten is actively recruiting global ecosystem partners, welcoming robotics manufacturers and university research teams to participate in joint research initiatives and frontier technology exploration. Tashan Technology and Openmind will provide tactile sensors and robot experimental platforms, supporting real tactile interaction data collection, reinforcement learning experiments, continuous learning research, collaborative publications, and academic exchanges.
The Robot Kindergarten has now entered its operational phase. Moving forward, Tashan Technology and Openmind will drive continued breakthroughs in key technologies and scenario validation across the seven research initiatives.