Project summary

This project develops a task planning solution using PDDL (Planning Domain Definition Language) and ROSPlan to automate restocking tasks in a retail store with a robot, addressing labor shortages in the retail sector due to an aging population in developed countries. The system enables a robot to restock products in a simulated store environment, determining the correct placement of items based on predefined store rules. The simulated environment includes various tables such as a stock table, refrigerated table, bin, and non-refrigerated table. A Python-based ontology is utilized to define product classes, with the store rules dictating where products, based on their characteristics like refrigeration needs or damage, should be placed. The initial knowledge base, established in PDDL files, outlines the robot's state and the environmental setup. The robot's task involves picking up products and placing them at designated locations. This process includes navigating between waypoints, picking items from the stock table, and placing them at their intended destinations, all governed by the store rules. ROSPlan interfaces with the knowledge base, updating it with product-specific predicates and executing a series of actions for each product instance. However, the planning and execution phase encountered technical challenges, particularly with the simulation in the ROS environment and issues executing the place action. While the generated plan aligned with the design and store rules, the simulation failed to effectively visualize the plan's execution. For future enhancements, the project suggests exploring alternative planners for more complex scenarios and resolving issues related to the place action interface. Additionally, customizing the simulated world further and using an alternative planner are recommended to enhance the system's efficiency and adaptability. The project illustrates the potential of using a knowledge representation and reasoning approach for automated retail store restocking, highlighting the need for further development for practical implementation.

Project documents

Report:

• Download PDF

Code implementation

Code for this blog:

• WEBSITE

Code to reproduce the results:

• RESULTS