Abstract
The use of drones is by now a standard resource in modern societies, enabling new inspection methods, cinematography, parcel delivery possibilities, as well as new personal
transportation paradigms. Notably, there are several multinational companies investing in automated delivery systems while the major players in aviation and aerospace industry are
also looking at new aerial urban mobility solutions using concept vehicles inspired on drones. Each of these applications face significant challenges in terms of regulatory,
technological, as well as scientific limitations related with the use of these vehicles, including endurance and efficiency, accurate perception of their surroundings, or in-flight scalable
cooperation with other vehicles. It is also worth noting the shift of the aerial transportation industry towards more individualized and clean energy solutions, to contribute to a more
sustainable development of cities.
Towards these goals, neither the exclusive use of a rotary-wing or a fixed-wing drone satisfy all expectations. While the former has notable maneuverability and can vertical take-off
and land (VTOL), the latter can be much more efficient to cover great distances for a given payload and has longer flight-times, at the cost of requiring dedicated runways or
complex structures for takeoff and land. A standard approach is to design a single hybrid vehicle that has VTOL capabilities, while still generating some lift from its wings when in
forward flight. This project challenges this approach, as the resulting hybrid drone is a compromise between maneuverability and efficiency. The proposed alternative strategy is to
design specialized vehicles for each task and have shuttle vehicles to enable launching and capture maneuvers in urban environments.
Thus, to mitigate the use of inefficient drones, this project will address the underlying scientific and technological challenges towards the use of shuttle drones to perform launch and
capture maneuvers of other vehicles. A first scenario is considered where the shuttle drones cooperate with the vehicles to be launched or captured, either relaying information about
their planned motion or actively synchronizing their motion to facilitate the maneuver. Addressing this scenario enables the design of specialized vehicles for a given task, for
instance, having fixed-wing drones with long endurance, while enabling the vertical take-off and landing of these vehicles with shuttle drones. A second scenario involves the launch
and capture of objects or other drones, that may not cooperate with the shuttle drones, either passive or actively. In this case, the shuttle drones may act as a security measure to
remove drones or objects from restricted areas that may actively avoid being captured, moved, and launched into safer areas.
Addressing these scenarios raises interesting challenges involving optimal and cooperative task allocation and planning of trajectories for a team of heterogeneous vehicles,
cooperative and distributed control for critical rendezvous maneuvers, as well as for the cooperative and distributed estimation of the motion of the shuttle drones, other vehicles,
and the surrounding environment. In addition, addressing the non-cooperative scenario will also involve dealing with estimation, control and planning strategies for maximizing the
chances of achieving the desired maneuver while minimizing the possibilities of the non-cooperative vehicle getting away. The research plan will start by addressing single vehicle
control and estimation strategies tailored for launch and capture maneuvers, and progressively evolve into cooperative control, estimation and planning for coordinated launch and
capture. The use of strategies building on differential games will be also considered in the non-cooperative scenario, going towards the use of techniques such as min-max model
predictive control. Both cooperative and noon-cooperative strategies will strive to incorporate learning-based control techniques, making use of the most recent results on deep
convolutional neural networks. As the projects aims at demonstrating the use of shuttle drones, the specification, design, and prototyping of the these vehicles will play a key role in
the development efforts. Nonetheless, the specification and design will also address example vehicles to be launched and captured.
The project team has a clear commitment towards the practical application of solid theoretical strategies, with a solid record of research contributions in the field of drone planning,
control, and navigation. The most relevant work includes SLAM and LiDAR-based control for autonomous rotorcraft, nonlinear model predictive control (MPC) strategies for
terrain following using rotorcraft and current rejection in marine vehicles, as well as nonlinear strategies based on Lyapunov theory or MPC for trajectory planning by single and
multiple coordinated vehicles.
Tasks and application scenarios
The proposed research plan is divided into the following tasks:
T1. Coordination and dissemination
T2. Specification, design, and vehicle operation
T3. Perception, navigation, and estimation strategies
T4. Cooperative task-allocation, planning and control
T5. Deep and non-cooperative control
T6. Datasets and experimental validation
Tasks T2 and T6 concentrate on the practical aspects of the project, whereas tasks T3 to T5 focus on the core of the project formed by the theoretical developments in perception
and estimation, motion control, planning and task allocation. Special emphasis will be placed on keeping a close interaction between the developments accomplished in tasks T3 to
T5 and the experimental implementation and testing of the resulting algorithms in task T6, as this task will play a fundamental role in validating the proposed solutions. Task T1 will
ensure a clear and smooth coordination of all institutions and research team, also launching an ambitious communication and dissemination strategy.
Although the scientific component, addressed in tasks T3, T4, and T5, are the foundations of the project, the design and experimental validation of the developed vehicles and
techniques, addressed in tasks T2 and T6, also have a key role for the success of the project, in particular, when considering possible knowledge transfer to the industry and society.
This practical component will be considered in three scenarios for launching and capture, which represent progressively more challenging problems, but also more interesting
possibilities of application and a greater degree of autonomy of the system:
S1. Launch and capture of passive vehicles or objects
S2. Launch and capture of actively cooperative vehicles
S3. Launch and capture of actively non-cooperative vehicles
These three scenarios will consider a single shuttle drone for the launching and capture maneuvers, yet applications of multiple shuttle drones will also be considered in the research
activities of the project. On the other hand, to validate the system with different types of target vehicles to be deployed or retrieved, two application scenarios will be considered:
(A1) launch and capture of a fixed-wind drone, and (A2) launch and capture of a light-weight marine vehicle, such as a glider or a surface vessel.