Composabl

Appearance

The Platform for Creating Skill-Based Intelligent Autonomous Agents ​

Composabl is the platform for creating industrial strength intelligent agents that make high-impact decisions in the real world.

With Composabl, you can convert expert knowledge about how a process works into a set of skills — discrete units of competence that allow the agent to make the right decision to control the system in a specific situation. These skills can be either programmed or learned through advanced AI techniques and then put together so that the agent knows how to perform in every part of the process and under any conditions. The platform then uses trial-and-error to train the agent in realistic scenarios until the agent can succeed at the task and outperform the alternatives.

Intelligent Autonomous Agents ​

Composab

  • Autopilot: An agent acts as an autopilot "brain" that avoids obstacles, flies to the destination, then lands the drone.
  • Operator: An agent guides human operators to adjust the speed of machines on a manufacturing line to reduce bottlenecks and maximize revenue.
  • Trader: An agent trades commodities on a market for maximum profit by reading trends and executing the best trading strategy at the right time.
  • Producer: An agent autonomously controls an extruder to make consumer product goods.
  • Scheduler: An agent schedules the production line in a factory to maximize the fulfillment of high-value orders.
  • Building Engineer: An agent controls the heating, ventilation, and air conditioning (HVAC) system in a building to reduce carbon footprint and save energy while still keeping occupants comfortable.
  • Assembler: An agent controls a robotic arm to pick and place parts on an assembly line.

Structure of Intelligent Autonomous Agents ​

Intelligent autonomous agents have structure, just like our brains. Different parts perform different functions.

Sensors take input from the environment and also provide feedback about the results of the agent’s actions. A perception layer contains modules that synthesize new information (like vision, hearing, trends, predictions) from the sensors. Skills decompose the task into modular sub-components and combine to fulfill the goals of the agent. The system then outputs decisions that take action on the system. During training, the agent gets feedback from a simulation environment to help it learn. Once it is deployed, it gets feedback from the real environment.

Engineers and process experts design agents using a method called Machine Teaching. Machine Teaching leverages insights about how humans learn new things to break down tasks into skills that the agent can acquire piece by piece. This allows intelligent agents to train quickly and efficiently, enables different technologies to control different parts of the process as appropriate, and makes AI systems accessible and explainable.

Learn More ​

This documentation details how to create agents using the Composabl SDK.

For more information and examples:

To learn more about machine teaching and how to design and build intelligent agents: