RESEARCH INVOLVEMENT
Developing Autonomy
ROBUST CONTROL OF AUTONOMOUS SURFACE VEHICLES
Autonomous Surface Vehicles (ASVs) operate in complex environments subject to external disturbances from wind, waves, and currents. Furthermore, mathematical model discrepancies to the real system, such as unmodeled dynamics or parameter inaccuracies, induce uncertainty. Thus, robust control techniques are required to overcome disturbances and uncertainty, increasing safety, efficiency, and accuracy.
The project targets motion planning in 2D complex but highly-structured environments. It aims at developing an automated motion planning research software toolbox that delivers highly performant trajectories at low computational cost, exhibits predictable behavior, and can shape the trajectories to specific application needs.
These requirements will be realized through a combination of three strategies:
(i) dividing trajectory into offline-optimized or expert-demonstrated application-specific motion primitives
(ii) using a fast optimal control solver only sparingly where most impactful for the overall optimality
(iii) a high-level router that safeguards the consistency of the solution and avoids deadlock and congestion.
COLLISION AVOIDANCE FOR AUTONOMOUS SURFACE VEHICLES
Based on the international RoboBoat competition, real-world applications of ASVs may have to deal with unstructured environments where the ASV has to avoid obstacles while doing a specific task. Hence, collision avoidance strategies are needed to ensure safety of the robot.
DEEP REINFORCEMENT LEARNING FOR AUTONOMOUS SURFACE VEHICLES
The complex nature of the ASV dynamics is an issue to develop mathematical models and design control systems. Since Deep Reinforcement Learning uses neural networks to learn a policy by interacting with its environment, how can they be applied for ASV autonomous behavior?
ASV-UAV HETEROGENEOUS TEAMS
Unmanned Aerial Vehicles may cooperate with ASVs for different marine applications such as surveillance and inspection. However, different tasks need to be accomplished including visual tracking, take-off, and landing.
ROBUST CONTROL OF UNMANNED UNDERWATER VEHICLES
Unmanned Underwater Vehicles (UUVs) are disturbed by the environmental conditions, particularly from underwater currents. Likewise, mathematical model inaccuracies and unknown dynamics create an uncertain system. Hence, robust control approaches able to counteract such perturbations are relevant.