Automated Test Selection for Simulation-based Testing of UAVs

Context

Unmanned aerial vehicles (UAVs), also known as drones, are acquiring increasing autonomy. With their commercial adoption, the problem of testing their safety requirements has become a critical concern. Simulation-based testing represents a fundamental practice for cost-effective testing of UAVs.

Motivation

Ensuring the reliability and safety of Unmanned Aerial Vehicles requires extensive simulation-based testing. Previous work [1] suggested that not all test cases contribute equally to assessing software quality, and executing “safe but uninformative” tests can lead to unnecessary resource consumption. While previous research has explored test optimization in self-driving cars domains, this challenge remains largely unaddressed for UAV simulation platforms.

Goal

This project aims to develop UAV-Scissor, a Machine Learning (ML)-based approach to improve the cost-effectiveness of simulation-based UAV testing. UAV-Scissor will analyze test cases before execution to predict their likelihood of revealing faults, allowing unnecessary tests to be skipped while maintaining high fault-detection capability.

The project will focus on designing ML models (or alternative approaches) for test case selection/classification (and/or prioritization), integrating UAV-Scissor into a UAV simulation environment (i.e., Aerialist [2,3]), and evaluating its effectiveness against traditional and random test selection methods.

Requirements

Recommended:

Reference(s)