Description:
In this edition of QuEST, Dr. Raj Sharma and Jasper Craig with the SAFE Autonomy Team in ACT3 will discuss their work on implementation of reinforcement learning (RL) in a simulated environment.
Key Moments in the video include:
Quadrotor Dynamics
Differential flatness
Linear Quadratic Controller (LQR)
Formation Consensus
Reinforcement Learning (RL) for Trajectory Control
Benefits of LQR and RL
Future Work
Audience questions:
Kinematics that you used for this control - does that change for battery-powered or certainly gas-powered weight of the craft changes? How do people model that, or is it necessary to model that?
Would there be a place in that matrix you were describing a moment ago to account for some of those things?
Does every agent or node have perfect knowledge of the leader?
What mechanisms were used to work with Kerianne’s group - Summer Faculty Fellowship? More