Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM67398
Volume:
10 |
9 |
Sep 2026
Pages:
1-8
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Science and Engineering
Vikrant University
Computer Science and Engineering
Vikrant University
The coordination of multiple unmanned aerial vehicles in shared low-altitude airspace requires decision support that can represent uncertainty, intercept unsafe actions, and explain interventions to an operator. This study presents X-AgentDeconflict, a two-dimensional simulation prototype that combines a heuristic multi-UAV navigation controller, Dempster-Shafer evidence fusion, a rule-based runtime safety shield, explanation generation, and an audit-oriented web dashboard. Wind magnitude, nearest-UAV distance, communication quality, and obstacle distance are converted into basic belief assignments and fused into a pignistic probability of safety. The implementation was inspected and then evaluated in 135 reproducible runs: three configurations were compared across 15 paired seeds with 10 UAVs, and the full configuration was evaluated at six fleet sizes from five to 30 UAVs with 15 seeds per size. Each run contained 200 steps at 0.1 s. No collision-threshold breach was observed in the 10-UAV comparison. Across the scalability study, however, breaches occurred in one of 15 trials at 20 UAVs, two of 15 at 25 UAVs, and five of 15 at 30 UAVs. Mean BetP (safe) declined from 0.907 at five UAVs to 0.827 at 30 UAVs, while the mean minimum separation declined from 104.86 m to 23.00 m. These findings show that the prototype supports uncertainty-aware monitoring and auditable intervention, but they do not establish a collision-free guarantee or validate a learned multi-agent reinforcement learning policy. The observed degradation at higher densities defines concrete requirements for predictive shielding, realistic wireless-network modelling, and three-dimensional validation.
KEYWORDS
X-AgentDeconflict demonstrates how Dempster-Shafer evidence fusion, a rule-based action shield, trace explanations, and an operator dashboard can be assembled into an auditable multi-UAV research prototype. A corrected 135-run evaluation found no collision-threshold breaches through 15 UAVs under the tested scenarios, but breaches appeared at higher fleet sizes and occurred in five of 15 trials at 30 UAVs. Mean BetP(safe) and minimum separation declined with density. At 10 UAVs, the shield and TBM configurations did not significantly improve separation relative to the heuristic baseline. The present evidence therefore supports feasibility and diagnostic value, not formal safety, real-world deployment, or MARL effectiveness. Future work should implement synchronous state updates, predictive multi-agent shielding, trained policies, realistic wireless co-simulation, longer adversarial scenarios, three-dimensional dynamics, and hardware-in-the-loop or flight validation.
1. Kopardekar P, Rios J, Prevot T, Johnson M, Jung J, Robinson JE: Unmanned Aircraft System Traffic Management Concept of Operations. NASA Ames Research Center, Moffett Field; 2016. 2. Federal Aviation Administration: Unmanned Aircraft Systems Traffic Management Concept of Operations, Version 2.0. FAA, Washington, DC; 2020. 3. European Commission: Commission Implementing Regulation (EU) 2021/664 on a regulatory framework for the U-space. Official Journal of the European Union. 2021. 4. Rashid T, Samvelyan M, de Witt CS, Farquhar G, Foerster J, Whiteson S: QMIX: monotonic value function factorisation for deep multi-agent reinforcement learning. Proceedings of the 35th International Conference on Machine Learning. 2018, 80:4295-4304. 5. Lowe R, Wu Y, Tamar A, Harb J, Abbeel P, Mordatch I: Multi-agent actor-critic for mixed cooperative-competitive environments. Advances in Neural Information Processing Systems. 2017, 30:6379-6390. 6. Yu C, Velu A, Vinitsky E, Gao J, Wang Y, Bayen A, Wu Y: The surprising effectiveness of PPO in cooperative multi-agent games. Advances in Neural Information Processing Systems. 2022, 35:24611-24624. 7. Chen YF, Liu M, Everett M, How JP: Decentralized non-communicating multiagent collision avoidance with deep reinforcement learning. 2017 IEEE International Conference on Robotics and Automation. IEEE, Singapore; 2017. 285-292. 10.1109/ICRA.2017.7989037 8. Alshiekh M, Bloem R, Ehlers R, Konighofer B, Niekum S, Topcu U: Safe reinforcement learning via shielding. Proceedings of the AAAI Conference on Artificial Intelligence. 2018, 32:2669-2678. 10.1609/aaai.v32i1.11797 9. Garcia J, Fernandez F: A comprehensive survey on safe reinforcement learning. Journal of Machine Learning Research. 2015, 16:1437-1480. 10. van den Berg J, Guy SJ, Lin M, Manocha D: Reciprocal n-body collision avoidance. Robotics Research. Springer, Berlin; 2011. 3-19. 10.1007/978-3-642-19457-3_1
Publication Date
9/23/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
1
Total views since publication
PDF Downloads
5
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...