On the Validation of a UAV Collision Avoidance System Developed by Model-Based Optimization: Challenges and a Tentative Partial Solution

Abstract

The development of the new generation of airborne collision avoidance system ACAS X adopts a model-based optimization approach, where the collision avoidance logic is automatically generated based on a probabilistic model and a set of preferences. It has the potential for safety benefits and shortening the development cycle, but it poses new challenges for safety assurance. In this paper, we introduce the new development process and explain its key ideas using a simple collision avoidance example. Based on this explanation, we analyze the challenges it poses to safety assurance, with a particular focus on system validation. We then propose a Genetic-Algorithm-based approach that can efficiently search for undesired situations to help the development and validation of the system. We introduce an open-source tool we have developed to support this approach and demonstrate it on searching for challenging situations for ACAS XU.