Titulo Estágio
UAVs Airspace Traffic Management via Artificial Neural Networks and Swarm Intelligence: Minimizing the Risk of Collision (in Cascade)
Áreas de especialidade
Engenharia de Software
Sistemas Inteligentes
Local do Estágio
DEI-FCTUC
Enquadramento
Unmanned aerial vehicles (UAVs) and Drones are aircrafts without a pilot boarding and can be controlled by the ground station pilot or an autonomous system. They have shown their potential in many military examples and now are getting acceptance in many different civilian applications such as deliveries, disaster management, rescue operations, geographic mapping, safety inspections, crop monitoring, hurricanes, and tornadoes monitoring, law enforcement and border control or visual inspection of structures.
The recent large-scale deployment of these devices has engendered rampant competition for airspace. In this regard, safety issues arise, namely, the likelihood of collisions increases due to several unplanned disturbances and hazards stemming from distinct causes: weather (exogenous), device software and hardware malfunction, security attacks, etc. Under crowded airspace, a particular collision is likely to yield a cascade of further collisions rendering navigation of this collective large-scale complex dynamical systems blatantly impractical. New navigation policies ought to be designed to adaptively foster safety while maintaining the mission(s) compliance.
More concretely, remark that each drone settles at a particular mission-driven region, also known as volume – i.e., the region in the airspace wherein drones are licensed to roam over actively. Safety margins are set between the volumes to hinder the possibility of collisions. If the safety margins are too slim, the risk of conflicts and collision (and resulting cascades thereof) is high. On the other hand, safety-margins conform to idle regions in airspace naturally yielding major navigation constraints and resulting in slimmer active regions for navigation. A trade-off should be met that rests on the following triptych: i) intrinsic attributes of the devices; ii) safety-driven distributed algorithms; iii) and the exogenous factors perturbing the course of the missions. In this internship, we will act on point ii), i.e., on designing distributed algorithms or distributed navigation policies to consistently grant safe navigation under demanding safety margins.
As we are studying a collective complex dynamical system, on the one hand, optimal navigation policies should be designed globally, i.e., leveraging on the whole vector of states and missions of the comprising drones. However, due to the large-scale dimension of the system, at each time window, drones are unable to aggregate the whole current state of the system but, can only base their current decisions on locally collected information, i.e., information exchanged with their closest peers. The nature of the information might span from the prospective whereabouts to intrinsic current attributes of the drones, e.g., malfunctioning, etc. Building on this local information exchange, we design the algorithms to a) detect/quantify the risk of collision (in cascade); b) drones optimal maneuvering to reduce/prevent the risk (if any) and advance in the mission.
Objetivo
The goal of this internship is twofold:
1) [Risk detection]. Given the current observed state of the complex (dynamical) system along with the spatial weather profile (windy, hot, etc.), device software/hardware failure and security attacks, devise Deep Learning-based tools to detect/quantify the risk of collisions consistently.
2) [Design of navigation policies]. Devise distributed algorithms based on swarm intelligence to i) minimize the collision risk while executing the mission; ii) manage emergency situations whereby risk has been detected. Swarm intelligence-based algorithms have been successfully applied to scenarios where collective tasks are on-demand under certain constraints (in our case, the safety margins).
In this internship, the student should study, propose, implement, and test methods for risk quantification and identification (via deep learning tools) and policy design to best mitigate the risk while complying with the mission of a large set of Drones.
More concretely, for the risk quantification module, the training set of the (deep convolutional) neural networks will be based on (synthetic) time-series reflecting the evolution of the swarm of drones obtained from proper mathematical models. We use the global imminent state to a cascade of collisions as input to the CNNs. The goal is to render the neural networks consistent over real world scenarios, i.e., that could consistently quantify the risk. Further, the student should explore swarm intelligence-based methods to design the optimal trajectories for the drones so to reduce the risk and advance in the mission.
In this internship, the student will be exposed to three hot topics of current research: Swarm intelligence, Deep learning, Airspace traffic management.
Plano de Trabalhos - Semestre 1
[Some tasks might overlap; M=Month]
T1 (M1-M2): Knowledge transfer and State of the art analysis (i.e., study the concepts behind UAV systems, U-space service, target level of safety in these systems, concept of separation minima, etc.).
T2 (M3): Identification of all hazards and issues that may have impact on safety (conflict and collision)
T3 (M4-M5): Propose methods for risk quantification and identification (via deep learning tools)
T4 (M6): Writing the Intermediate Report.
Plano de Trabalhos - Semestre 2
[Some tasks might overlap; M=Month]
T5 (M7-M8): Implementation of the proposed methods for risk quantification and identification (via deep learning tools)
T6 (M9-M10): Policy design to best mitigate the risk while complying with the mission of a large set of Drones
T7 (M11): Writing the thesis.
Condições
The selected student will be integrated in the Software and Systems Engineering (SSE) group of CISUC and the work will be carried out in the facilities of the Department of Informatics Engineering at the University of Coimbra (CISUC - Software and Systems Engineering Group), where a work place and necessary computer resources will be provided.
Note: there is a possibility of having a 6-month scholarship.
Type of scholarship: Bolsa de Investigação para Licenciado (Renewable).
Value of the scholarship: 752,38€.
Observações
Please contact the advisors for any question or clarification needed.
Advisors: Naghmeh Ivaki (naghmeh@dei.uc.pt) and Augusto Santos (augustosantos@dei.uc.pt).
Orientador
Naghmeh Ivaki / Augusto Santos
naghmeh@dei.uc.pt 📩