![]() Implementing these algorithms can involve using frameworks and packages like ROS, Gazebo, Python, C++, OpenCV, PCL, and PyTorch. #Air force research laboratory softwareThis work could involve conducting research in nonlinear state estimation using local feedback, reinforcement learning, or adaptive control for multi-agent systems, and implementing algorithms related to these topics in software (e.g., Gazebo) and on test platforms (e.g., quadrotors). The focus of this work is to investigate recent results in literature that have demonstrated methods to estimate the dynamics of a system, the uncertainty associated with those dynamics, and the state of the system, and to develop multi-agent policies that are robust to intermittence in feedback. #Air force research laboratory updateApproaches to solve these problems often use reinforcement learning and other adaptive methods to determine stabilizing policies, but often assume the state of the system is known or use local feedback (e.g., cameras and lidar) in nonlinear state estimators that do not update system dynamics models. ![]() In these scenarios, the solution for the optimal guidance and control policies is unknown and nonlinear. Contact mentorĪdaptive Multi-Agent Cooperation Subject to Intermittent MeasurementsĪutonomous multi-agent tasks in uncertain and contested environments require robustness to intermittent feedback and faults in vehicles. Depending on the background of the scholar, the project could involve: developing reinforcement learning or adaptive control theoretical results, implementing algorithms/software on test platforms (e.g., quadrotors or UAVs) and testing, extending existing work to a multi-agent domain, or software testing/development. The current goal of this research is to make decisions (e.g., between trajectory-following and hover) or improve performance based on this real-time data. This project will involve learning from, building upon, and accelerating currently on-going work related to the development of online real-time data-driven estimation and control methods for flight control systems. Contact mentorĪdaptive Dynamical Learning and Control for Flight Vehicles The specifics of the project will adjust based off of the skillset of the intern but it will cover developing either new capabilities within our current scene generation software in C++ or implementing new features in a 3D modeling software using python. The core of this project is developing new scene generation capabilities for use in developing more complex, realistic scenes efficiently. To test these sensors, scene generation is used to generate artificial worlds for the sensor to see. Digital testing of objects can uncover errors and issues before the system ever transitions to the field for real-world testing. When developing new types of sensors, it is important to test the hardware and software components in controlled environments. 3D Modeling and Scene GenerationĪcademic Level: Lower-level Undergraduate, Upper-level Undergraduate To contact the mentor, use the link included at the conclusion of each project description. Scholars are encouraged to contact any mentors whose projects they find of interest. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |