Evaluation of Probabilistic Conflict Detection and Resolution for BVLOS UAS Under Surveillance and Communication Constraints

Abstract

Beyond visual line-of-sight unmanned aircraft systems are being deployed in low-altitude airspace for inspection, logistics, and monitoring, where heterogeneity of platforms and infrastructures intensifies the need for systematic conflict management. Existing separation assurance concepts often assume reliable cooperative surveillance and continuous communication, conditions that are not guaranteed in fragmented coverage, mixed equipage, or contested environments. Under such conditions, deterministic conflict detection and rule-based resolution may be either too conservative or operationally fragile, motivating explicit treatment of probabilistic uncertainty, delayed state information, and intermittent command and control links. This paper examines probabilistic conflict detection and resolution strategies for multiple vehicles operating beyond visual line-of-sight subject to surveillance errors, correlated track losses, communication latency, packet loss, and asynchronous decision making. A unified modeling framework is constructed in which navigation, sensing, and communication artifacts contribute to stochastic relative state estimates used to trigger conflict alerts and synthesize resolution maneuvers. Several algorithmic variants are evaluated, including prediction-based alerting with bounded risk, decentralized chance-constrained trajectory adjustment, and policies that adapt to context-dependent loss-of-link probabilities. The study quantifies how surveillance quality, communication availability, and algorithm design interact to influence conflict probability, alert volume, trajectory deviation, and residual mid-air collision risk across representative operational scenarios. The analysis aims to describe performance and trade-offs without assuming idealized infrastructures, providing a basis for parameterizing risk-based separation requirements and for assessing the feasibility of scalable beyond visual line-of-sight operations under realistic technical constraints.