intrinsic-dev/aic

What it solves

This toolkit provides the necessary infrastructure for participants of the AI for Industry Challenge, a competition focused on solving high-impact robotics and manufacturing problems, specifically the task of commanding a robot to insert cables.

How it works

The toolkit is split into two main parts: an evaluation component and a participant model component. The evaluation component (provided by organizers) handles the simulation environment (Gazebo), robot control, sensor fusion, and scoring. Participants develop a ROS 2 node (the aic_model) that processes sensor data—such as camera images and force/torque measurements—and outputs motion commands to the robot.

Who it’s for

Roboticists and developers interested in creating AI-driven control policies for industrial manufacturing tasks.

Highlights

• ROS 2 Integration: Built on ROS 2 (specifically Kilted Kaiju) for standardized robot communication.
• Simulation-First: Includes a full simulation environment with aic_engine for trial orchestration and validation.
• Policy Framework: Provides a ready-to-use aic_model framework to handle ROS 2 boilerplate, allowing developers to focus on the Python policy logic.
• Comprehensive Tooling: Includes adapters, controllers, and example policies to accelerate development.