[Paper Review] Distributed Mission Planning of Complex Tasks for Heterogeneous Multi-Robot Teams

Barbara Arbanas, Tamara Petrović+1 more|arXiv (Cornell University)|Sep 21, 2021

Modular Robots and Swarm Intelligence4 citations

TL;DR

This paper proposes a distributed, multi-stage optimization method for planning complex missions in heterogeneous multi-robot teams, using a hierarchical task model and a two-phase approach: first, a heuristic search identifies multiple viable mission decompositions; second, a distributed genetic algorithm with mimetism performs multi-objective task allocation and scheduling. The method adapts to robot availability and optimization criteria, demonstrating effective trade-offs between makespan and energy cost in a simulated greenhouse environment.

ABSTRACT

In this paper, we propose a distributed multi-stage optimization method for planning complex missions for heterogeneous multi-robot teams. This class of problems involves tasks that can be executed in different ways and are associated with cross-schedule dependencies that constrain the schedules of the different robots in the system. The proposed approach involves a multi-objective heuristic search of the mission, represented as a hierarchical tree that defines the mission goal. This procedure outputs several favorable ways to fulfill the mission, which directly feed into the next stage of the method. We propose a distributed metaheuristic based on evolutionary computation to allocate tasks and generate schedules for the set of chosen decompositions. The method is evaluated in a simulation setup of an automated greenhouse use case, where we demonstrate the method's ability to adapt the planning strategy depending on the available robots and the given optimization criteria.

Motivation & Objective

• To address the challenge of planning complex, heterogeneous multi-robot missions with cross-schedule dependencies and multiple execution paths.
• To develop a domain-independent, distributed approach that enables adaptive mission planning based on available robots and optimization criteria.
• To improve robustness and scalability in mission planning by decoupling task decomposition from allocation and scheduling.
• To evaluate the method’s ability to balance competing objectives—such as time efficiency and energy consumption—under varying robot configurations.
• To demonstrate the method’s adaptability in real-world scenarios, such as automated greenhouses, through simulation.

Proposed method

• The mission is modeled as a hierarchical task tree inspired by TÆMS, enabling structured representation of complex, decomposable tasks.
• A fast heuristic tree search algorithm explores alternative task decomposition paths to generate multiple favorable mission execution strategies.
• A distributed multi-objective genetic algorithm with mimetism and knowledge sharing performs task allocation and scheduling across robots.
• The algorithm optimizes for both makespan and energy cost, with customizable weighting based on user-defined criteria.
• Knowledge sharing between distributed agents enhances convergence and solution quality in the optimization process.
• The method supports heterogeneous robots with distinct capabilities, including mobile manipulators, stationary manipulators, and UGVs, and accounts for task dependencies and cross-schedule constraints.

Experimental results

Research questions

• RQ1 How can a distributed, multi-stage optimization framework effectively balance competing objectives—such as mission duration and energy consumption—in heterogeneous multi-robot systems?
• RQ2 To what extent can a heuristic-based mission decomposition selection improve planning efficiency and solution diversity in complex task hierarchies?
• RQ3 How does the integration of mimetism and knowledge sharing in a distributed genetic algorithm enhance solution quality and convergence in task allocation and scheduling?
• RQ4 Can the proposed method dynamically adapt mission execution strategies based on available robot types and user-defined optimization priorities?
• RQ5 How does the method perform under varying robot configurations, such as mobile vs. stationary manipulators, in real-world application scenarios?

Key findings

• The method successfully generated multiple viable mission decompositions, with the best solution balancing makespan and cost based on user-defined weights.
• Using one stationary manipulator and two light UGVs reduced energy consumption by a factor of 16 compared to two mobile manipulators, though mission duration increased by 57%.
• When cost was prioritized (setup 3), the system selected energy-efficient task execution paths aligned with the left branch of the mission tree.
• When speed was prioritized (setup 5), the system favored faster execution via mobile manipulators, selecting the right branch of the mission tree.
• With equal importance of cost and makespan (setup 4), the solution balanced task distribution across both high-efficiency and high-speed execution paths.
• The simulation results confirm the method’s adaptability to different robot configurations and optimization criteria, demonstrating effective trade-off management in complex mission planning.