Anas Hossini

Design of a predictive failure model for wastewater treatment plants in the Essonne region (France ), aiming to anticipate system and component failures using data-driven approaches. ∠Based on these predictions, adaptive maintenance strategies are defined using Deep Reinforcement Learning (DRL) to optimize decision-making, reduce downtime, and improve overall system reliability.

Design of predictive maintenance models for smart buildings:

∠ Design of a predictive failure model in a complex system (smart building)

∠ Design of a cooperative multi-agent reinforcement learning model to define maintenance strategies

∠ Optimization of maintenance costs under quality of service constraints using a simulated annealing metaheuristic integrating different neighborhoods

∠ Performance comparison between metaheuristics (simulated annealing, tabu search) and reinforcement learning algorithms (DQN, A2C, PPO)

∠ Proposal of an innovative hybrid approach integrating RL and OR to simultaneously optimize maintenance costs and quality of service

Reinforcement learning for dynamic 5G resource management:

∠ Simulation of a 5G network using a queue management approach

∠ Design of a reinforcement learning model for resource management within a 5G network architecture (trade-off between bandwidth and CPU)