Autobot

Autobot represents a versatile line of self learning dual arm robots designed for automating both industrial and household tasks. At its core, Autobot leverages a lightweight, locally pretrained AI multimodal system, serving as its cognitive and locomotion center. This AI system seamlessly interacts with perception and motion sensors to detect various objects and execute actions.

Here’s how Autobot operates:

Task Mapping and Action Curation: 

The cognitive center simplifies complex tasks by breaking them down into manageable steps. It maps the complex objects used in the process to their core shapes, such as 3D rectangles or circles, and then generates a set of curated actions tailored to each object. For example, if the task involves picking up an object from the side and then turning right, the cognitive center determines these necessary actions.

Action Segregation and Perception: 

Autobot categorizes actions into two main types: movement actions and checking actions. Movement actions involve physical motion, such as picking up an object or adjusting its position. Checking actions verify the state of the task, ensuring accuracy and completeness.

Locomotion and Execution: 

Once the actions are segregated, Autobot sends a request to the locomotion center to calculate precise angles and directions for movement actions based on the localization information of objects. This informs how the necessary actions should be performed. For example, if Autobot needs to pick up an object, it calculates the optimal approach angle and coordinates. Once computed, these actions are transmitted to the control system and executed in a serialized manner.

Error Handling: 

Autobot’s control system continuously monitors its actions. If an error occurs (e.g., an object slips during pickup), the control system prompts Autobot to request guidance on how to proceed. The entire process then repeats, from finding the slipped object to recalculating locomotive movements.