Partial and dynamic reconfiguration provides a relevant new dimension to design efficient parallel embedded systems. However, due to the encasing complexity of such systems, ensuring the consistency and parallelism management at runtime is still a key challenge. So architecture models and design methodology are required to allow for efficient component reuse and hardware reconfiguration management. The proposed approach is inspired from the well-known component based models used in software applications development. Our model is based on membranes wrapping the systems components. The objective is to improve design productivity and ensure consistency by managing context switching and storage using modular distributed hardware controllers. These membranes are distributed and optimized with the aim to design self-adaptive systems by allowing dynamic changes in parallelism degree and contexts migration.
In this work we deal with a combinatorial optimization problem applied to a real and
concrete case. It consists in designing automated containers for storage of medicine boxes
in a pharmacy. Thus, the pharmacy has limited storage space (undivided containers), in
which the flow of medicines must be handled automatically by a robot. The goal is to use
the available storage space in the best possible way.
The resolution that I propose for this problem consists of two main steps:
• Dividing the containers (’3 D’ separation: width, depth and height)
• Management of the boxes flow (arrivals and exits)
The first step consists in the classic problem of cutting, for which there are several resolution
approaches in the literature. However, the existing solutions, related to ’3D ’
aspects, are not suitable for the resolution of my problem. We are therefore inspired by
the literature for instances ’2 D ’ by adapting mainly greedy heuristics. Thus, many of
these approaches have been adapted to the 3D context. They are also implemented and
compared, in terms of performance and execution time.
In the second stage of the approach that we propose, the medicines flow has been
modeled as a problem of multi-period multi-dimensional knapsack with objects removal.
The goal is to maximize the total profit of the pharmacy. This new problem differs from
classical knapsack problems found in the literature by taking into account the time of
object release. Thus we have proposed several heuristics to solve this particular problem.