Can we get our robots to "think" the same way, with a little "voice" that tells them what to do when they are presented with a complex task? We developed a system that we call the Hierarchical Interactive Robot (Hi Robot) that allows us to incorporate vision-language-action (VLA) models, such as π0, into a two-level inference process. π0 serves as the instinctual, reactive "System 1" that can perform well-practiced tasks, and a high-level semantic vision-language model (VLM) plays the role of "System 2," reasoning through complex tasks and language interactions by "talking to itself." This System 2 high-level policy quite literally emulates that little voice, telling the robot how to break up complex tasks into intermediate steps
Can we get our robots to "think" the same way, with a little "voice" that tells them what to do when they are presented with a complex task? We developed a system that we call the Hierarchical Interactive Robot (Hi Robot) that allows us to incorporate vision-language-action (VLA) models, such as π0, into a two-level inference process. π0 serves as the instinctual, reactive "System 1" that can perform well-practiced tasks, and a high-level semantic vision-language model (VLM) plays the role of "System 2," reasoning through complex tasks and language interactions by "talking to itself." This System 2 high-level policy quite literally emulates that little voice, telling the robot how to break up complex tasks into intermediate steps
