Angebot 1
How Far Can Deep Reinforcement Learning Scale for Operational Planning?
Englische Zusammenfassung
This IDP investigates how the scalability of deep reinforcement learning-based operational planning changes with problem size. Students will implement a DRL training framework for comparable CPU- and GPU-based experiments, vary problem and system parameters, measure computational and performance metrics, and evaluate bottlenecks and GPU acceleration.
Überblick
- Projektdetails
- How Far Can Deep Reinforcement Learning Scale for Operational Planning?
Textstelle · Seite 1
- “How Far Can Deep Reinforcement Learning Scale for Operational Planning?” Seite 1
- Organisation
- Im PDF nicht angegeben
- Projektformate
- IDP
Textstelle · Seite 1
- “IDP” Seite 1
- IDP
- Studienabschlüsse
- Master
Textstelle · Seite 1
- “The thesis is suitable for master’s students in Computer Science.” Seite 1
- Master
- Projektziel
- Investigate the scalability boundary of current DRL-based planning methods.
Textstelle · Seite 1
- “The project will analyze how training time, memory usage, sample efficiency, and policy quality change as the size of the operational planning problem increases.” Seite 1
Themen und Aufgaben
- Themen
- Deep reinforcement learning (DRL)
Textstelle · Seite 1
- “Deep Reinforcement Learning (DRL) has shown strong potential for solving complex sequential decision-making problems in operations management” Seite 1
- operational planning
Textstelle · Seite 1
- “realistic, large-scale operational planning problems” Seite 1
- scalability of DRL-based planning methods
Textstelle · Seite 1
- “The objective of this thesis is to investigate the scalability boundary of current DRL-based planning methods.” Seite 1
- GPU-based simulation and policy training
Textstelle · Seite 1
- “A particular focus will be placed on understanding whether GPU-based simulation and policy training can provide the speedups required for practically relevant large-scale applications.” Seite 1
- Deep reinforcement learning (DRL)
- Anwendungsbereiche
- inventory control
Textstelle · Seite 1
- “including inventory control, production planning, routing, scheduling, and resource allocation.” Seite 1
- production planning
Textstelle · Seite 1
- “including inventory control, production planning, routing, scheduling, and resource allocation.” Seite 1
- routing
Textstelle · Seite 1
- “including inventory control, production planning, routing, scheduling, and resource allocation.” Seite 1
- scheduling
Textstelle · Seite 1
- “including inventory control, production planning, routing, scheduling, and resource allocation.” Seite 1
- resource allocation
Textstelle · Seite 1
- “including inventory control, production planning, routing, scheduling, and resource allocation.” Seite 1
- inventory control
- Methoden und Werkzeuge
- Deep Reinforcement Learning (DRL)
Textstelle · Seite 1
- “Deep Reinforcement Learning (DRL) has shown strong potential for solving complex sequential decision-making problems in operations management” Seite 1
- CPU- and GPU-based experiments
Textstelle · Seite 1
- “implement a DRL training framework that can run comparable CPU- and GPU-based experiments.” Seite 1
- GPU-based simulation
Textstelle · Seite 1
- “whether GPU-based simulation and policy training can provide the speedups required for practically relevant large-scale applications.” Seite 1
- policy training
Textstelle · Seite 1
- “whether GPU-based simulation and policy training can provide the speedups required for practically relevant large-scale applications.” Seite 1
- Deep Reinforcement Learning (DRL)
- Aufgaben
- Define scalable operational planning problems
Textstelle · Seite 1
- “Define scalable operational planning problems” Seite 1
- Implement a DRL training framework for comparable CPU- and GPU-based experiments
Textstelle · Seite 1
- “implement a DRL training framework that can run comparable CPU- and GPU-based experiments.” Seite 1
- Systematically vary problem size, batch size, network size, and simulator design
Textstelle · Seite 1
- “Systematically vary problem size, batch size, network size, and simulator design” Seite 1
- Measure training time, memory usage, sample throughput, and performance
Textstelle · Seite 1
- “to measure training time, memory usage, sample throughput, and performance.” Seite 1
- Identify computational bottlenecks
Textstelle · Seite 1
- “Identify the main computational bottlenecks” Seite 1
- Evaluate whether GPU acceleration enables practically relevant scaling
Textstelle · Seite 1
- “evaluate whether GPU acceleration enables practically relevant scaling.” Seite 1
- Define scalable operational planning problems
- Art der Arbeit
- Erwartete Ergebnisse
- DRL training framework
Textstelle · Seite 1
- “implement a DRL training framework that can run comparable CPU- and GPU-based experiments.” Seite 1
- DRL training framework
Voraussetzungen
- Erforderliche Kenntnisse
- Ability to work independently
Textstelle · Seite 1
- “The ability to work independently, as well as analytical skills, are required.” Seite 1
- Analytical skills
Textstelle · Seite 1
- “The ability to work independently, as well as analytical skills, are required.” Seite 1
- Profound knowledge of Python
Textstelle · Seite 1
- “Profound knowledge of Python (preferably JAX / PyTorch) is required.” Seite 1
- Ability to work independently
- Wünschenswerte Kenntnisse
- Geeignete Studienrichtungen
- Computer Science
Textstelle · Seite 1
- “The thesis is suitable for master’s students in Computer Science.” Seite 1
- Computer Science
- Programmierung im Projekt
- Zentral
Textstelle · Seite 1
- “implement a DRL training framework that can run comparable CPU- and GPU-based experiments.” Seite 1
- Programmierkenntnisse als Voraussetzung
- Erforderlich
Textstelle · Seite 1
- “Profound knowledge of Python (preferably JAX / PyTorch) is required.” Seite 1
Rahmenbedingungen
- Teamgröße
- Im PDF nicht angegeben
- Dauer
- Im PDF nicht angegeben
- Beginn
- October 2026Normalisiertes Datum: 2026-10
Textstelle · Seite 1
- “Earliest begin: October 2026” Seite 1
- Bewerbungsfrist
- Im PDF nicht angegeben
- Arbeitsort
- Im PDF nicht angegeben
- Arbeitsmodus
- Im PDF nicht angegeben
- Arbeitssprache
- Im PDF nicht angegeben
Lernen und Betreuung
- Lernmöglichkeiten
- Im PDF nicht angegeben
- Angebotene Betreuung
- Im PDF nicht angegeben
Bewerbung und Kontakt
- Kontaktpersonen
- Patrick Helm
Textstelle · Seite 1
- “Supervisor: Patrick Helm” Seite 1
- logtheses.log@mgt.tum.de
Textstelle · Seite 1
- “Email with curriculum vitae and transcript of records to logtheses.log@mgt.tum.de” Seite 1
- Patrick Helm
- Bewerbung
- Email with curriculum vitae and transcript of records to logtheses.log@mgt.tum.de
Textstelle · Seite 1
- “Email with curriculum vitae and transcript of records to logtheses.log@mgt.tum.de” Seite 1
- Weitere Informationen
- Im PDF nicht angegeben