Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Active Inference-Driven World Modeling for Adaptive UAV Swarm Trajectory Design

Kaleem Arshid, Ali Krayani, Lucio Marcenaro, David Martin Gomez, Carlo Regazzoni

TL;DR

This work addresses adaptive UAV swarm trajectory design under uncertainty by combining Active Inference with a hierarchical, symbolic World Model learned from GA-RF expert demonstrations. The offline phase builds Mission, Route, and Motion dictionaries that encode stereotyped plans, while online inference minimizes divergence from reference distributions to realize adaptive partitioning, ordering, and motion generation. A state-estimation module (EKF) ensures smooth, collision-free navigation. Simulation results show faster convergence, higher stability, and safer navigation than a modified Q-Learning baseline, demonstrating scalability and cognitive grounding.

Abstract

This paper proposes an Active Inference-based framework for autonomous trajectory design in UAV swarms. The method integrates probabilistic reasoning and self-learning to enable distributed mission allocation, route ordering, and motion planning. Expert trajectories generated using a Genetic Algorithm with Repulsion Forces (GA-RF) are employed to train a hierarchical World Model capturing swarm behavior across mission, route, and motion levels. During online operation, UAVs infer actions by minimizing divergence between current beliefs and model-predicted states, enabling adaptive responses to dynamic environments. Simulation results show faster convergence, higher stability, and safer navigation than Q-Learning, demonstrating the scalability and cognitive grounding of the proposed framework for intelligent UAV swarm control.

Active Inference-Driven World Modeling for Adaptive UAV Swarm Trajectory Design

TL;DR

This work addresses adaptive UAV swarm trajectory design under uncertainty by combining Active Inference with a hierarchical, symbolic World Model learned from GA-RF expert demonstrations. The offline phase builds Mission, Route, and Motion dictionaries that encode stereotyped plans, while online inference minimizes divergence from reference distributions to realize adaptive partitioning, ordering, and motion generation. A state-estimation module (EKF) ensures smooth, collision-free navigation. Simulation results show faster convergence, higher stability, and safer navigation than a modified Q-Learning baseline, demonstrating scalability and cognitive grounding.

Abstract

This paper proposes an Active Inference-based framework for autonomous trajectory design in UAV swarms. The method integrates probabilistic reasoning and self-learning to enable distributed mission allocation, route ordering, and motion planning. Expert trajectories generated using a Genetic Algorithm with Repulsion Forces (GA-RF) are employed to train a hierarchical World Model capturing swarm behavior across mission, route, and motion levels. During online operation, UAVs infer actions by minimizing divergence between current beliefs and model-predicted states, enabling adaptive responses to dynamic environments. Simulation results show faster convergence, higher stability, and safer navigation than Q-Learning, demonstrating the scalability and cognitive grounding of the proposed framework for intelligent UAV swarm control.
Paper Structure (8 sections, 11 equations, 5 figures)

This paper contains 8 sections, 11 equations, 5 figures.

Table of Contents

  1. Introduction
  2. System and Problem Setup
  3. Proposed Framework
  4. Expert Demonstrations via GA--RF
  5. Hierarchical Symbolic World Model
  6. Online Decision-Making via Active Inference
  7. Results
  8. Conclusion

Figures (5)

  • Figure 1: Hierarchical behavior: (a) mission division and routing; (b) local motion via learned motion words.
  • Figure 2: Adaptive replanning to a new target: (a) surprise; (b) belief update and corrected path.
  • Figure 3: EKF-assisted correction and obstacle avoidance: (a) prediction/correction; (b) static (black) and dynamic (pink) obstacles.
  • Figure 4: Trajectories: Active Inference vs. Modified QL.
  • Figure 5: Quantitative comparison: (a) time; (b) distance.