Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bayesian Active Inference for Intelligent UAV Anti-Jamming and Adaptive Trajectory Planning

Ali Krayani, Seyedeh Fatemeh Sadati, Lucio Marcenaro, Carlo Regazzoni

TL;DR

The paper tackles resilient UAV operation under adversarial jamming by marrying offline expert demonstrations with online Bayesian Active Inference in a hierarchical world model. It introduces three dictionaries within a Generalized Dynamic Bayesian Network to jointly encode symbolic planning, low-level motion, and SINR-driven perception, enabling online inference that minimizes abnormality relative to expert plans. The approach yields near-expert performance in terms of interference reduction and mission efficiency, with robust generalization across jammer densities and accurate jammer localization (RMSE < 20 m). This framework offers a principled, scalable path for secure and adaptive UAV trajectories in contested wireless environments.

Abstract

This paper proposes a hierarchical trajectory planning framework for UAVs operating under adversarial jamming conditions. Leveraging Bayesian Active Inference, the approach combines expert-generated demonstrations with probabilistic generative modeling to encode high-level symbolic planning, low-level motion policies, and wireless signal feedback. During deployment, the UAV performs online inference to anticipate interference, localize jammers, and adapt its trajectory accordingly, without prior knowledge of jammer locations. Simulation results demonstrate that the proposed method achieves near-expert performance, significantly reducing communication interference and mission cost compared to model-free reinforcement learning baselines, while maintaining robust generalization in dynamic environments.

Bayesian Active Inference for Intelligent UAV Anti-Jamming and Adaptive Trajectory Planning

TL;DR

The paper tackles resilient UAV operation under adversarial jamming by marrying offline expert demonstrations with online Bayesian Active Inference in a hierarchical world model. It introduces three dictionaries within a Generalized Dynamic Bayesian Network to jointly encode symbolic planning, low-level motion, and SINR-driven perception, enabling online inference that minimizes abnormality relative to expert plans. The approach yields near-expert performance in terms of interference reduction and mission efficiency, with robust generalization across jammer densities and accurate jammer localization (RMSE < 20 m). This framework offers a principled, scalable path for secure and adaptive UAV trajectories in contested wireless environments.

Abstract

This paper proposes a hierarchical trajectory planning framework for UAVs operating under adversarial jamming conditions. Leveraging Bayesian Active Inference, the approach combines expert-generated demonstrations with probabilistic generative modeling to encode high-level symbolic planning, low-level motion policies, and wireless signal feedback. During deployment, the UAV performs online inference to anticipate interference, localize jammers, and adapt its trajectory accordingly, without prior knowledge of jammer locations. Simulation results demonstrate that the proposed method achieves near-expert performance, significantly reducing communication interference and mission cost compared to model-free reinforcement learning baselines, while maintaining robust generalization in dynamic environments.

Paper Structure

This paper contains 13 sections, 22 equations, 7 figures.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. Proposed Anti-Jamming UAV Trajectory Design Based on Bayesian Active Inference
  4. Expert Demonstration Generation
  5. Hierarchical World Model Construction
  6. Bayesian Active Inference for Online Trajectory Adaptation
  7. Online High-Level Planning
  8. Online Low-Level Planning and Signal-Driven Jamming Response
  9. Results and Discussion
  10. System Setup
  11. Online Trajectory Adaptation
  12. Performance Evaluation
  13. Conclusion

Figures (7)

  • Figure 1: Schematic representing the main steps of the offline process to build the World Model.
  • Figure 2: Schematic of the main steps in the Active Inference online process.
  • Figure 3: Dictionaries structured as GDBNs: (a) Dictionary 1, (b) Dictionary 2, (c) Dictionary 3.
  • Figure 4: Online trajectory adaptation under jamming using Bayesian Active Inference.
  • Figure 5: Performance vs. expert and Q-learning across region counts.
  • ...and 2 more figures