Testing and Performance Evaluation Methods for UAS Detect-and-Avoid Using Simulated and Live Encounter Scenarios

Abstract

 As UAS operations expand in density and complexity, detect-and-avoid capabilities are expected to provide an acceptable level of risk with respect to mid-air collisions, while remaining compatible with existing traffic and scalable to emerging concepts such as dense urban operations. Evaluating detect-and-avoid performance under both nominal and off-nominal encounter geometries requires carefully constructed methodologies that integrate high-fidelity simulation, controlled live-flight experiments, and rigorous statistical analysis. This paper examines test and evaluation constructs for detect-and-avoid systems with emphasis on harmonizing simulated encounter sets and live scenarios, enabling quantitative assessment of detection performance, maneuver guidance, interoperability, and residual risk. The discussion focuses on systematic strategies for constructing encounter models, defining operational scenarios, specifying performance metrics, and integrating measurement uncertainty, without relying on a single environment or data source. A particular emphasis is placed on traceability between modeled encounters and operational use cases, including those characterized by sparse surveillance, heterogeneous equipage, and mixed levels of automation. The paper outlines approaches that connect algorithmic behavior with safety-relevant indicators such as conflict rate, loss-of-well-clear frequency, and collision probability, in a way that allows incremental validation and refinement. The resulting framework enables transparent interpretation of detect-and-avoid performance across simulated and live encounter campaigns while remaining adaptable to different system architectures and operational concepts.