Intelligent infrastructure maintenance with deep reinforcement learning

Home

Principal Investigators

• Prof. Dr. Daniel Straub, Daniel, Engineering Risk Analysis Group, Departments of Civil and Environmental Engineering

• Prof. Dr.-Ing. habil. Alois Christian Knoll, Chair of Robotics, Artificial Intelligence and Real-time Systems, School of Computation, Information and Technology

PhD Students

• Carmen Buliga, M.Sc., Engineering Risk Analysis Group, Departments of Civil and Environmental Engineering
• Daniel Hettegger M.Sc., Chair of Robotics, Artificial Intelligence and Real-time Systems, School of Computation, Information and Technology

Post Docs

• Florian Walter, Dr. rer. nat., Chair of Robotics, Artificial Intelligence and Real-time Systems, School of Computation, Information and Technology

Funding

TUM Georg Nemetschek Institute Artificial Intelligence for the Built World

Project summary

Maintenance of civil infrastructure systems, including transportation, energy and water networks, comes at a major cost to society. At present, planning of maintenance actions (which includes inspections) is based mainly on a combination of fixed rules and ad-hoc optimization. As a result, there is a large potential for improved planning. However, optimal planning is challenging, because there is significant uncertainty on the condition of the infrastructure components and because there are a large number of parameters, interactions and constraints, including reliability and safety requirements.

The planning of infrastructure maintenance belongs to the realm of sequential decision making, in which decisions are made in a sequential manner under uncertainty. Such optimization problems are challenging. However, in recent years, deep reinforcement learning (DRL) has become increasingly powerful in solving this class of problems. DRL utilizes deep neural networks to describe large state spaces and appears to be well suited for application to infrastructure maintenance planning.

This project aims to make a significant step towards AI-supported maintenance planning by (a) setting up formal descriptions of infrastructure maintenance planning as a sequential decision problem and (b) developing tailored DRL algorithms for identifying optimal maintenance strategies. To pave the way for the scientific advances to find their way into practice, we will collaborate with infrastructure owners and operators and showcase our methods on gas and road networks.

Current events

• Workshop on “Deep reinforcement learning for infrastructure maintenance planning: Trends, opportunities and challenges”