An Optimal Task Planning and Agent-aware Allocation Algorithm in Collaborative Tasks Combining with PDDL and POPF

TL;DR

This paper presents a flexible framework combining PDDL, POPF, and a task allocation algorithm to generate optimal collaborative plans for robots and humans in industrial settings, enhancing adaptability over fixed plans.

Contribution

It introduces a novel task allocation algorithm considering practical factors and integrates it with PDDL and POPF for optimal planning in collaborative tasks.

Findings

Successful execution of assembly tasks with robots and humans.

Demonstrated improved flexibility and safety in planning.

Validated the approach through real-world experiments.

Abstract

Industry 4.0 proposes the integration of artificial intelligence (AI) into manufacturing and other industries to create smart collaborative systems which enhance efficiency. The aim of this paper is to develop a flexible and adaptive framework to generate optimal plans for collaborative robots and human workers to replace rigid, hard-coded production line plans in industrial scenarios. This will be achieved by integrating the Planning Domain Definition Language (PDDL), Partial Order Planning Forwards (POPF) task planner, and a task allocation algorithm. The task allocation algorithm proposed in this paper generates a cost function for general actions in the industrial scenario, such as PICK, PLACE, and MOVE, by considering practical factors such as feasibility, reachability, safety, and cooperation level for both robots and human agents. The actions and costs will then be translated into a language understandable by the planning system using PDDL and fed into POPF solver to generate an optimal action plan. In the end, experiments are conducted where assembly tasks are executed by a collaborative system with two manipulators and a human worker to test the feasibility of the theory proposed in this paper.