Presenter Email
akbasm@erau.edu
Location
Mori Hosseini Student Union Events Center (Bldg #610) – Rooms 165 B/C
Start Date
3-3-2020 10:45 AM
End Date
3-3-2020 12:00 PM
Submission Type
Presentation
Other Topic Area
Autonomous Aerial Vehicles
Keywords
Autonomous Aerial Vehicles, Validation, Testing, Modeling and Simulation, Formal Methods
Loading...
Abstract
Autonomous aerial vehicles (AAV) have the potential to have market disruptions for various industries such as ground delivery and aerial transportation. Hence, the USAF has called for increased level of autonomy and rapid progress is being made in artificial intelligence (AI) engines, complex and non-deterministic system components, which are at the core of the autonomous aerial platforms. Traditional testing and validation methods fall short of satisfying the requirement of testing such complex systems. Therefore, to achieve highly or fully autonomous capabilities, a major leap forward in the validation is required. Otherwise, the full capabilities of autonomy will not be realized. The key challenges are the localization of problems, development of object models for perception and the creation of a safety measure. A similar challenge exists in ground autonomous vehicles (AVs), where there is a significant investment in recent years. However, there are important differences in the environmental and regulatory conditions between these two domains. In this paper, we present a validation framework that uses modeling and simulation and formal methods for solving the issues in the validation of AAVs. We define an abstraction stack using techniques such as separation of concerns, constrained pseudo-random test generation, coverage driven testing and functional assertions. We focus on the identification of edge cases and using test cases build a coverage matrix to determine the criteria for success. The system aims to assess the creation of an evolving safety measure and a licensing structure.
Validation Framework for Autonomous Aerial Vehicles
Mori Hosseini Student Union Events Center (Bldg #610) – Rooms 165 B/C
Autonomous aerial vehicles (AAV) have the potential to have market disruptions for various industries such as ground delivery and aerial transportation. Hence, the USAF has called for increased level of autonomy and rapid progress is being made in artificial intelligence (AI) engines, complex and non-deterministic system components, which are at the core of the autonomous aerial platforms. Traditional testing and validation methods fall short of satisfying the requirement of testing such complex systems. Therefore, to achieve highly or fully autonomous capabilities, a major leap forward in the validation is required. Otherwise, the full capabilities of autonomy will not be realized. The key challenges are the localization of problems, development of object models for perception and the creation of a safety measure. A similar challenge exists in ground autonomous vehicles (AVs), where there is a significant investment in recent years. However, there are important differences in the environmental and regulatory conditions between these two domains. In this paper, we present a validation framework that uses modeling and simulation and formal methods for solving the issues in the validation of AAVs. We define an abstraction stack using techniques such as separation of concerns, constrained pseudo-random test generation, coverage driven testing and functional assertions. We focus on the identification of edge cases and using test cases build a coverage matrix to determine the criteria for success. The system aims to assess the creation of an evolving safety measure and a licensing structure.
Comments
Presented during Concurrent Session 6A: Urban Air Mobility