Autonomous Work Machines

The development of reliable and high-performance perception systems is crucial for robotic systems and autonomous vehicles—especially under challenging outdoor conditions. These systems must process incomplete or noisy data, handle limited visibility or adverse weather conditions, and approximate the dynamic range and sensitivity of human senses as closely as possible.

To achieve this, data fusion methods, context-based selection processes, and innovative sensor technologies are essential. By combining robust perception with the ability to understand and interpret the context and structure of a scene, we develop smarter systems for a wide range of applications.

Since work machines operate in dynamically changing and unstructured environments, the risk of collisions is particularly high, potentially causing significant damage to property and serious injuries. Therefore, their manual operation requires years of experience and great caution.

We develop, integrate, and test assistance functions—up to fully autonomous operation of your work machine—to ensure effective, robust, and safe task execution. In combination with our per-ception systems, we detect and process changes in the environment and dynamically adjust task and motion planning in real time to ensure collision-free and efficient operation.

With adaptive control algorithms, we achieve maximum precision even under changing system conditions, such as load fluctuations, wear, or aging.

By combining adaptive drive control, dynamic perception, and intelligent planning methods, we ensure that the right actions are executed at the right time.

In the development of complex autonomous systems, model-based and data-driven subprocesses are interconnected. Verification directly on a prototype comes with numerous challenging obstacles and is also costly, time-consuming, and risky. Therefore, simulation-based verification and further development play a crucial role in our development process.

Physically and visually realistic digital twins also provide a valuable opportunity for operators to test future assistance functions in a virtual environment.

We are capable of replicating the dynamic behaviour of the work machine and its interaction with the environment—whether navigating rough terrain, adapting to various weather conditions, or interacting with humans—using realistic simulation environments. This allows potential issues to be identified and resolved before deployment in the real world. To achieve this, we combine model-based and data-driven modelling with generative AI algorithms in a hybrid approach.

Our simulations enable rapid innovation cycles, reduce costs and development time, and provide a safe testing environment. Additionally, we use digital twins as training simulators to optimally prepare operators for real-world applications.

For autonomous systems, safety and reliability are the core criteria for deployment: Avoiding standstills and accidents is of the utmost importance. To minimise the occurrence of either, we employ complementary sensor technologies combined with intelligent sensor fusion algorithms. The majority principle allows for reliable emergency recognition even in harsh environments. 

Our control concepts are designed to be reliable without sacrificing efficiency. Innovative verification and validation methods allow us to systematically test the algorithms under various environmental disturbances and inherent system parameter uncertainties. This enables meaningful statistical evaluation of our algorithms’ reliability.