Robust Algorithms for Complex Autonomous Robot Systems
Abstract
This project addresses a concern of the broad AI and autonomy communities: as autonomy is pushed into complex and unstructured environments, the vulnerabilities of sensing, processing, and decision-making must be better understood and addressed. This effort focuses on mitigating autonomous robot system (ARS) vulnerabilities that can be potentially exploited. A comprehensive study of possible disruptions to ARSs will be performed, followed by development of robust autonomy algorithms that mitigate these disruptions. Analyses will be undertaken to develop a deep understanding of how uncertainty in ARSs affects the autonomy framework: sensors to software. This effort will culminate in novel ARS architectures for executing robust autonomy in military-relevant scenarios. The primary objective is to provide hands-on, military-relevant education for graduate and undergraduate students in ARSs, Robot Operating System (ROS), simulation environments (e.g., Gazebo, Unreal Engine, and AirSim), and robust autonomy algorithms. The secondary objectives are to develop a fundamental understanding of the behavior of ARSs in the presence of disruptions and uncertainty, and to investigate sensor, hardware, and instrumentation considerations with respect to robust autonomy algorithms to establish a deep understanding of the real-world implications on current and future military ARSs. Our students will undertake four tasks in their investigation of robust ARS sensing and algorithms. Relevant research literature on robust autonomy, methods, and algorithms will be cataloged, creating a “playbook” of robust autonomy methods and potential uncertainty-inducing disruptions. Simulated and real-world development, testing, and validation of robust autonomy algorithms will be performed. ROS and Gazebo will serve as the simulation software environment. Promising methods from simulations will be demonstrated in real-world experiments using various ARSs. A flexible open-source codebase of autonomy-supporting functionality that enables agile development of robust autonomy algorithms will be created. Last, the students will develop novel robust autonomy algorithms and adaptive autonomy functionalities, testing these in military-relevant scenarios.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 12, 2025
- Source ID
- N001742310003
Entities
People
- Timothy C Havens
Organizations
- Michigan Technological University
- United States Navy