Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Constrained multi-objective optimization for multi-UAV planning

Cristian Ramirez-Atencia, David Camacho

TL;DR

The paper addresses constrained multi-objective planning for multi-UAV missions, modeling a comprehensive problem with task allocation, temporal relations, flight profiles, sensors, and GCS assignments. It introduces MOEA-CSP, a hybrid approach that integrates a constraint satisfaction problem into an NSGA-II based evolutionary framework, using six-allele encoding to represent assignments, orders, profiles, and sensor usage. Through experiments on six increasing-complexity missions, it demonstrates how adding variables expands the search space while incorporating CSP-based constraints can improve convergence, albeit with higher per-generation cost; constraints also reduce the feasible solution space and influence Pareto-front quality. The work provides practical insights for deploying constrained MOEA-based planners in real-world multi-UAV operations and points to decision-support enhancements to rank and filter Pareto-optimal plans.

Abstract

Over the last decade, developments in unmanned aerial vehicles (UAVs) has greatly increased, and they are being used in many fields including surveillance, crisis management or automated mission planning. This last field implies the search of plans for missions with multiple tasks, UAVs and ground control stations; and the optimization of several objectives, including makespan, fuel consumption or cost, among others. In this work, this problem has been solved using a multi-objective evolutionary algorithm combined with a constraint satisfaction problem model, which is used in the fitness function of the algorithm. The algorithm has been tested on several missions of increasing complexity, and the computational complexity of the different element considered in the missions has been studied.

Constrained multi-objective optimization for multi-UAV planning

TL;DR

The paper addresses constrained multi-objective planning for multi-UAV missions, modeling a comprehensive problem with task allocation, temporal relations, flight profiles, sensors, and GCS assignments. It introduces MOEA-CSP, a hybrid approach that integrates a constraint satisfaction problem into an NSGA-II based evolutionary framework, using six-allele encoding to represent assignments, orders, profiles, and sensor usage. Through experiments on six increasing-complexity missions, it demonstrates how adding variables expands the search space while incorporating CSP-based constraints can improve convergence, albeit with higher per-generation cost; constraints also reduce the feasible solution space and influence Pareto-front quality. The work provides practical insights for deploying constrained MOEA-based planners in real-world multi-UAV operations and points to decision-support enhancements to rank and filter Pareto-optimal plans.

Abstract

Over the last decade, developments in unmanned aerial vehicles (UAVs) has greatly increased, and they are being used in many fields including surveillance, crisis management or automated mission planning. This last field implies the search of plans for missions with multiple tasks, UAVs and ground control stations; and the optimization of several objectives, including makespan, fuel consumption or cost, among others. In this work, this problem has been solved using a multi-objective evolutionary algorithm combined with a constraint satisfaction problem model, which is used in the fitness function of the algorithm. The algorithm has been tested on several missions of increasing complexity, and the computational complexity of the different element considered in the missions has been studied.
Paper Structure (27 sections, 23 equations, 8 figures, 4 tables, 1 algorithm)

This paper contains 27 sections, 23 equations, 8 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. State of the art
  3. Mission Planning
  4. Constraint Satisfaction Problems
  5. Evolutionary Algorithms and Multi-Objective Optimization
  6. Description of the Multi-UAV Mission Planning Problem
  7. Task allocation
  8. Temporal relations
  9. Flight Profiles
  10. Sensors
  11. GCS assignments
  12. Checking constraints
  13. Path Constraints
  14. LOS constraints
  15. Optimization variables
  16. ...and 12 more sections

Figures (8)

  • Figure 1: NSGA-II overview of multiobjective strategies.
  • Figure 2: Climb/Descend situation in path computing.
  • Figure 3: Example of an individual that represents a possible solution for a problem with 5 tasks, 3 UAVs and 2 GCSs.
  • Figure 4: Mission Scenarios considered.
  • Figure 5: Hypervolume, number of solutions, number of generations and runtime obtained with NSGA-II for MPPs considering different approaches using different variables of the .
  • ...and 3 more figures