Automation Lab

A new car in the manufacturing area. — Volvo Cars is one of LiU:s partners

The Automation Lab develops methods and technologies for making engineering and production processes more automated, intelligent and adaptive.

Our research combines artificial intelligence, design automation, digital twins, robotics, simulation and structured engineering knowledge. The aim is to develop systems that can support 鈥� and increasingly perform 鈥� complex engineering tasks.

A central challenge is that engineering knowledge today is distributed across documents, databases, CAD models, simulations, production systems and human expertise. We investigate how this information can be made accessible, structured and machine-actionable so that both humans and intelligent systems can use it throughout the engineering lifecycle.

Our research has its foundation in Design och Production Automation, where engineering knowledge is formalised in models, rules and algorithms to automate repetitive engineering activities. Advances in artificial intelligence now make it possible to take the next step 鈥� from automating individual tasks towards connected and increasingly autonomous engineering systems.

Our main research areas include:

鈥� Agentic engineering systems
鈥� Structured engineering knowledge and trustworthy AI
鈥� Digital twins and adaptive automation
鈥� Engineering information access and downstream automation
鈥� AI-supported design, simulation and manufacturing

Research areas

Agentic engineering systems

We develop modular AI-agent architectures capable of interacting with engineering information, software tools and computational models.

Agents can retrieve information, analyse structured data, use engineering tools, coordinate specialised functions and execute multi-step workflows.

Our research investigates how such systems can support activities ranging from information retrieval and engineering analysis to product development, production preparation, robotics and quality assurance.

An important research question is how responsibility should be distributed between language models, deterministic software, structured knowledge and human engineers.

Structured engineering knowledge and trustworthy AI

Engineering organisations generate large amounts of information, but much of it remains difficult for both humans and machines to access.

Requirements, standards, drawings, reports, production instructions, simulation results and engineering decisions are commonly distributed across different systems and document formats.

We investigate methods for transforming this information into structured engineering knowledge while maintaining relationships to evidence, requirements, revisions, decisions and provenance.

This enables AI systems to operate on explicit information rather than relying solely on the internal reasoning or context of a language model.

A particular focus is the development of architectures for traceable and verifiable AI-supported engineering, including concepts such as the Relational Control Plane.

Digital twins and adaptive automation

Digital twins provide a connection between physical systems, engineering models and operational data.

Our research investigates digital twins that combine simulation, structured databases, sensors, computer vision and artificial intelligence.

We are particularly interested in digital twins that evolve from passive representations of physical systems into active components of intelligent automation.

Applications include adaptive robot systems, production planning, manufacturing processes and environments where conditions cannot be completely predicted in advance.

Engineering information and downstream automation

A significant amount of engineering time is spent finding, interpreting and transferring information between different activities and systems.

We investigate how AI and structured knowledge can make engineering information easier to access while simultaneously enabling automation further downstream.

For example, information originating in requirements, technical documentation or product models may ultimately support automated:

design and configuration
simulation and analysis
production preparation
manufacturing
robot programming
verification
quality assurance
regulatory compliance

The objective is not simply faster information retrieval, but to create information structures that can be used directly by subsequent engineering processes.

AI-supported design and manufacturing

We develop and apply computational methods for automating design and manufacturing activities.

Our research includes machine learning, optimization, physics-informed methods, computer vision and generative approaches together with established engineering methods such as CAD, CAE and simulation.

Applications range from automated geometry generation and engineering analysis to manufacturing planning, fixture design, production systems and industrial robotics.

Industry challenges

Two central challenges are making engineering knowledge machine-actionable and moving from isolated automation towards intelligent, connected engineering processes.

Fragmented engineering knowledge

Industrial organisations generate large amounts of information throughout product development and production. Requirements, standards, drawings, CAD models, simulations, production instructions, quality data and experience-based knowledge are often stored in different systems and formats.

A significant part of engineering work therefore still consists of searching for, interpreting, verifying and transferring information between different activities and organisations. In many cases, the information is accessible to humans but is not structured in a way that allows digital systems to interpret and use it directly.

The development of large language models and generative AI creates new possibilities for accessing previously difficult-to-use unstructured information. However, in engineering applications, it is not sufficient for an AI system to generate a plausible answer. Information must also be connected to requirements, sources, revisions, decisions and other technical relationships.

We therefore investigate how AI can be combined with structured databases, knowledge representations and explicit verification mechanisms. The goal is to make engineering knowledge accessible, traceable and machine-actionable.

Once information has this structure, it can also be used for more than information retrieval. The same knowledge can enable downstream automation in areas such as design, simulation, production preparation, manufacturing, robot programming, verification and quality assurance.

From isolated automation to intelligent engineering systems

Engineering automation has traditionally focused on well-defined and repetitive tasks. Parametric CAD, Knowledge-Based Engineering, optimisation and automated computational workflows have enabled substantial improvements in many parts of product development.

Modern engineering processes, however, consist of a large number of interconnected activities using different models, software tools and information sources. A design change may, for example, affect simulation, production preparation, manufacturing, robot programming, quality assurance and documentation.

Artificial intelligence and agentic systems create new possibilities for coordinating such processes. AI agents can retrieve information, use specialised software tools, analyse structured data and execute workflows consisting of multiple steps.

At the same time, our research is based on the principle that a language model should not be responsible for the entire engineering process. Instead, we combine different technologies according to their strengths:

鈥� Language models for interpretation, interaction and handling ambiguity
鈥� AI agents for planning and coordinating workflows
鈥� Structured knowledge and databases for facts, relationships, requirements and traceability
鈥� CAD, simulation and other engineering tools for specialised calculations and analyses
鈥� Verification mechanisms for checking results against defined requirements
鈥� Human engineers for objectives, experience, judgement and responsibility

Digital twins play an important role in this development by creating a connection between physical systems, engineering models and operational data. By combining digital twins with sensors, computer vision, simulation and AI, engineering systems can also adapt to changes in the physical environment.

The long-term objective is to move from digital tools that support isolated activities towards intelligent engineering systems in which information, models, AI and humans can collaborate throughout the development and production process.