In this project we want to solve blocks-world problem. We have a table with four blocks in different colors and five spots where we can place the blocks (four of them are already taken in the given state). To this state we’ll call the initial state.
Our goal is, with autonomous agent, to solve the problem. Meaning, change the blocks’ order to an order given by the user in the minimal amount of steps.
This problem is a famous problem in the automated planning world. Today, there is no known way to teach an autonomous agent to pick up objects efficiently. The problem we deal with in the project is a simpler version of this problem.
In the project the autonomous agent will solve the problem base on Sense-Think-Act methodology.
The solution stages are:
Sense – the agent need to deduct the position of the blocks from the world using Baxter robot cameras. The agent will process the image received from the cameras to a center of masses vector that will include the initial state.
Think – writing a planning problem using PDDL language. We’ll use this tool to write a problem that includes the initial state of the world, the goal state (received as input from user) and which action the agent can take to move from the initial state to the goal state. The solution to this problem is an action plan that tell the agent how to move the blocks to reach the goal state from the initial state, in minimal amount of steps.
Act – the agent will execute the given orders, with a movement plan, and will rearrange the blocks to reach the wanted goal state.
In conclusion, the agent solves the problem using the three stages described, meaning the problem defined in the project resolves as desired.