Intelligent Working Machines

Whether in manufacturing, construction, forestry, or logistics, robots and autonomous machines will increasingly support people at work by taking over heavy, repetitive, or hazardous tasks. At the same time, they can help meet growing demands for safety, efficiency, and sustainability, while also addressing the rising shortage of skilled workers across many industries.

To achieve this, intelligent machines require a broad range of capabilities that humans naturally take for granted but must first be “taught” to machines – from perception through advanced sensor systems to motion planning, control, and verifiable decision-making. Automation only becomes practical in real-world applications when all these components work together reliably under changing environmental conditions. Artificial intelligence plays a crucial role in this process – modern robotics essentially brings AI into motion.

In recent years, the AIT has achieved major advances in close collaboration with universities such as TU Wien and Tufts University, as well as industrial partners including Palfinger, Liebherr, and Künz.

Six research breakthroughs representing key building blocks for autonomous working machines will now be presented at the world’s leading robotics conference, IEEE International Conference on Robotics and Automation 2026, taking place in Vienna from June 1–5, 2026.

• Robust 3D Object Perception

Under the name PIRATR, researchers at AIT have developed an end-to-end AI system for three-dimensional object recognition based on laser scan data. The system goes beyond conventional object detection by not only identifying an object’s position and orientation but also describing variable object states, such as the opening angle of a crane gripper. These capabilities provide a crucial foundation for robust automation processes and safe interaction between autonomous working machines.

• Precise Placement of Heavy Components Using Cranes

In real-world applications, the accurate handling and placement of loads with cranes is often complicated by pendulum-like swinging motions. A new system now enables predictive control based on camera-based position estimation. Combined with collision-safe motion planning, it allows autonomous pickup and precise placement of components while avoiding obstacles with high accuracy and stable motion execution. This is particularly important in dynamic and cluttered construction environments.

• How Robotic Systems Learn Flexibly While Remaining Safe

Outside traditional industrial environments, autonomous machines require flexibility and real-time responsiveness. Working machines must adapt to changing conditions and disturbances without violating safety distances or exceeding motion, force, and workspace limits.

Previous learning and optimization methods could not fully satisfy these requirements simultaneously. A new method called “SafeFlowMPC,” developed at TU Wien with participation from AIT, overcomes this challenge in real time. A learning model generates motion proposals that are continuously verified and, if necessary, adjusted by an algorithm to ensure safety constraints are always maintained – even in dynamic situations.

• Safety in Autonomous Timber Loading Operations

In highly unstructured environments such as forests, timber-loading cranes must fulfill two critical requirements simultaneously: avoiding collisions and actively damping load oscillations.

Researchers have developed the first collision-free, oscillation-damping model predictive controller and implemented it on a timber-loading crane. The system reacts to its environment in real time: it can avoid obstacles dynamically, replan operations when conditions change, continue collision-free operation under disturbances, and automatically stop the machine if no safe maneuver is possible.

• How Robots Learn From Repetitive Processes

In autonomous vehicle racing, an ideal racing line is initially planned and then systematically improved lap after lap using feedback from the control system. Deviations from the planned trajectory are not treated as disturbances but rather as indicators of locally challenging conditions. This creates an adaptive map of permissible accelerations that continuously improves lap times.

The same principle can be applied to autonomous working machines performing repetitive tasks. When a system learns from execution deviations, its movements become more robust and efficient over time – even under changing conditions.

• Reliable Autonomy for Long Task Sequences

Many autonomous systems can already perform individual tasks effectively. The real challenge arises when tasks consist of many consecutive steps — such as picking up, transporting, positioning, and placing objects – where every step must build correctly on the previous one.

In these cases, planning, intermediate states, and long-term consistency become critical. Among several possible approaches, a so-called neuro-symbolic architecture proved to be both more reliable and more energy efficient. This approach combines symbolic task planning – including sequences, rules, and verification steps – with learned low-level skills that execute individual motions robustly. With transparent task logic and reliable motion skills, autonomous systems can operate dependably in everyday environments.

World Premiere for an Autonomous Crane

These research results are directly integrated into the further development of autonomous machines at the AIT Large-Scale Robotics Lab, with the goal of continuously improving both safety and performance. The focus lies on technologies that prove themselves in real-world environments: robust against disturbances, safe in operation, and efficient in execution.

Visitors can experience these technologies firsthand at the “Festival der Roboter” on May 30–31, 2026, at Vienna’s Karlsplatz. For the first time in public, a fully autonomous robotic crane developed jointly by AIT, TU Wien, and Palfinger will be demonstrated live.

By combining imaging sensors, artificial intelligence, systems theory, and physical domain knowledge, the crane can autonomously load and unload timber logs onto trucks. Such autonomous and assistive systems support humans intelligently in complex tasks while improving safety, productivity, and competitiveness.

“Robotics is a key technology for our industrial future and is increasingly becoming part of everyday life. At the AIT Austrian Institute of Technology, we are actively shaping this development – from autonomous systems to AI-supported assistance technologies that support people in their work and enable entirely new applications. Our goal is to bring cutting-edge research rapidly into practical use together with our partners. In this way, we create solutions that make processes more efficient, increase safety, and sustainably strengthen the innovative power and competitiveness of our industries,” explains Andreas Kugi.

Festival der Roboter, May 30–31, 2026 – Karlsplatz

Detailed program: https://www.roboterinwien.com