Rapid Deployment of Autonomous Systems

Developers and researchers can often achieve successful systems in laboratory settings when working with autonomous systems. However, the systems frequently fail when they are introduced to real­world conditions. System owners must then dedicate significant time to adapting their systems to the real­world conditions present in the deployment environment. In the best­case scenario, this additional time merely increases costs. However, for some systems, such as emergency systems, a delay in deployment will render the system non­functional. Therefore, advancing research in rapid deployment could unlock new opportunities for autonomous systems. Autonomous systems often initially fail when deployed despite their successful functioning in laboratory environments. This discrepancy arises from the infinite complexity and unknowns of the deployment environment compared to the controlled conditions in the laboratory environment. These discrepancies, which can be called the deployment gap, represent the challenges faced when transitioning a system from a controlled laboratory environment to the complex real world. The process of deploying a system is to overcome this gap. system can bridge the deployment gap by being robust enough to withstand unknown challenges or being highly adaptive, meaning it is easy to understand and modify, enabling quick modifications to overcome the gap. For the most effective approach, both robustness and adaptivity should be used. Analyzing the time and access available in the deployment environment can help make a targeted effort toward the most effective deployment in a given scenario. While adaptivity tends to be more effective, it demands greater deployment access. Robustness is more universally applicable but is often harder to achieve. Examining the literature on other researchers’ strategies to bridge the deployment gap can determine how to achieve robustness and adaptivity in autonomous systems. The literature notes that researchers attempt to close the deployment gap by testing their systems and avoiding complex solutions. Through testing, researchers uncover unknown variables that may impact their system and use the new knowledge of their system to improve it. Employing simpler systems can minimize complex interactions with the environment, making it easier to work with and more robust. While most of the literature focuses on the testing aspect, the two concepts work together. Despite the known methods for closing the deployment gap, rapid deployment is still an issue, as seen in robotic competitions, such as the Mohamed Bin Zayed International Robotics Challenge (MBZIRC), where most teams fail to deploy their systems in time. This project has conducted several smaller projects to validate the effectiveness of utilizing adaptivity and robustness to achieve rapid deployment. The systems produced by the largest projects have been additionally validated through their participation in the MBZIRC. The projects described in this thesis are the following:
• A Domain­Specific Language was built to interface with and describe the movements of a manipulator robot. This language displayed rapid deployability through adaptiveness and robustness.
• The Cut­and­Paste method for generating data was expanded to enable faster data generation, specifically for deployment scenarios. The project demonstrated how data augmentation methods can enhance testing speed and understanding of specific systems. Rapid Deployment of Autonomous Systems.
• Modular task programming was used to quickly generate mission scripts for a combined mobile and manipulate,r creating a highly adaptive system. This system was used to participate in the MBZIRC and achieve top positions, validating the adaptive approach for rapid deployment.
• A study was conducted examining the effectiveness of simulation to improve deployment speed. The study revealed that while simulation is an effective tool for testing and enhancing deployment through robustness, it does require additional development time.
• A specific long­range fiducial marker was designed to expand the use cases for fiducial markers. The project showed that fiducial markers can improve deployment speed through better robustness.
• A self­contained navigation system was designed and deployed in the 2024 MBZIRC. The deployment showed the robustness achieved by utilizing self­contained systems.
• Additionally, the thesis describes lessons learned about deployment obtained during participation in the MBZIRC that were not directly related to any project.
Finally, the insights from the projects and literature research have been distilled into a three­step process. This process guides developers in designing systems with a focused approach to rapid deployment.