Dr. Warren Dixon

Newton C. Ebaugh Professor and Department Chair
Mechanical & Aerospace Engineering, UF

Dr. Warren Dixon and his students are focused on the development of theoretical methods motivated by systems that exhibit uncertain nonlinear behaviors. His approach to compensate for uncertainties in the dynamics is the development of adaptive and learning control methods. Recent projects have focused on the development of control systems for robotic agents where sensor information is intermittent.

Dr. Dixon’s Naval Surface Warfare Center project Collaborative Autonomous Maritime Vehicles for MCM Missions focuses on collaborative unmanned surface vehicles (USV) and autonomous underwater vehicles (AUV) for mine countermeasure (MCM) missions and aims to optimize time, energy, and sensing resources to develop a percentage clearance estimate. This proposal investigates computationally efficient optimal control solutions for a network of MCM AUVs despite constrained sensing/communication. The specific project objectives include:

  • Development of artificial learning methods where the computational burden is divided among cooperating assets and the optimal control problem is divided into spatial regions
  • Development of switched systems methods that generate timing conditions and stability guarantees for the networking of the AUVs and USVs can lead to undesirable responses, including loss of an asset, wasted time/energy to re-identify
  • Development of differential games that determine the intermittent interaction between the AUVs and USVs

Dr. Dixon’s Office of Naval Research project Approximate Optimal Online Continuous-Time Path-Planning Methods is focused on developing path-planning methods to operate in the complex littoral environment that optimally react to uncertain conditions. The specific project objectives include:

  • Advance the state of the art in model-based reinforcement learning (MBRL) to develop real-time optimal path-planning
  • Improve efficiency and accuracy of MBRL path-planning by concurrently determining optimality near the vehicle’s current state and destination
  • Evaluate these novel approaches to path-planning in relevant naval mission scenarios and compare the results to existing methods