October 4, 2021
PhD Seminars I
Alesia Herasimenka (McTao)
Impact of optical properties on the controlability of solar sails
Solar sails offer a propellant-less solution to achieve interplanetary transfers, planet escapes, and de-orbiting maneuvers by leveraging on solar radiation pressure (SRP). Despite very few solar sail missions were launched, the temptation to use SRP as an inexhaustible source of propulsion attracted the interest of researchers since decades, and several contributions on the guidance and control of solar sails are available. But the question of controllability remains unsolved.
Orbital control with SRP is challenging because the sail can not generate a force toward the direction of the Sun. Therefore, classical approach to study controllability can not be applied. Controllability means that the sail can move around its initial position. Thus, the fundamental question is: can we modify the orbit of the sail in any desirable way? We tackle this question by formulating a convex semi-infinite optimization problem.
The results show that the sail is locally controllable under a minimal condition on its reflectivity, and minimum requirement for solar sail design is formulated. This requirement can be used for solar sail technical design, as well as for defining its life time due to the phenomenon of optical degradation.
Jules Mabon (Ayana)
Combining point processes and CNNs for detection in remote sensing
In the case of remotely sensed images, detecting small targets such as vehicles is a arduous task when using frameworks developped for natural images (where objects of interest occupy a good portion of the frame). In such a context, when objects are small and their density is high, we use object interaction priors within a point process simulation. The data term of this point process is learned with a neural network, thus avoiding the handcrafting of a specific detection term.