Where does the use of autonomous systems bring the most benefit?

Autonomous vehicle systems, whether on land, at sea or in the air, can be applied in different areas due to their diverse characteristics. Applications range from drones for inspection or monitoring tasks to field robots for ecological pest control, to autonomous ships for freight or passenger transport. Therefore, the work of the department focuses on the following question: In which application scenarios can the extraordinary potential of autonomous systems be used in a meaningful way?

The considerations are based on the premise of making processes more resource-efficient, more flexible and more productive, as well as actively designing solutions for current and future challenges. Against the background that countless applications seem conceivable, it is also vital to examine whether using an autonomous system generates an economic and/or social added value and can thus be seen as an increase in efficiency or further development of existing processes.

Autonomous systems in port logistics

Due to their diverse characteristics, autonomous vehicle systems can be applied in different areas – here the example of port logistics

Which autonomous system is suitable for which application?

After the researchers at TITUS have determined the corresponding deployment scenario, they must then select the appropriate basic system for the defined mission. In order to do this, they analyse the corresponding influencing factors and mission-specific parameters.

The complex topic of “city logistics”, for example, can benefit from the targeted use of UAVs (Unmanned Aerial Vehicles), ground-based vehicle systems, and also autonomously operating inland vessels. Accordingly, the environment and available or required infrastructures (see Department of Infrastructures), as well as legal and regulatory requirements, must also be considered in the analysis.

What do autonomous systems need to be equipped with in order to be able to move?

In order to navigate safely through a dynamically changing environment, autonomous vehicles must have the following capabilities: fast and precise detection of their environment (perception), knowledge of their own location (localisation), and planning of a collision-free route to a defined destination (path planning).

Reliable data is crucial in ensuring these capabilities. Hence, the use of different sensor technologies, as well as the preparation and processing of the generated sensor data with the help of suitable algorithms, is a central task of the department. The main challenges are latencies, information gaps, measurement errors, and accuracy. Suppose one thinks of use under poor visibility conditions or in areas with diffuse structures (e.g. off-road applications). In that case, the question arises as to how safe autonomous operation with measurement uncertainties is possible at all. Here, for example, sensor fusion methods offer possible solutions.

In addition to these basic functions, which every system needs for autonomous operation, system-side adaptations are required depending on the application scenario. Thus, special hardware and software components must be additionally developed, both to fulfil the specific mission and for integration into a larger overall system.

Making autonomous systems capable of action through machine learning methods

Autonomous systems must be able to act in a situation-appropriate and anticipatory manner, especially under conditions that were previously unknown to them. They have to analyse the current situation within a concise time frame and draw conclusions for adapting their behaviour. Machine learning technologies are used for this. Based on given decision-making and evaluation problems, patterns or rules are derived from corresponding training data, which can later be adapted to similar problems. The quality of the reaction of the autonomous system depends very much on the quality of the training data. The highest possible quality of training data should be achieved by using different types of data or fused data. At the same time, simulation methods help to extend the training data with synthetic data.

How do data processing and control interact?

The autonomous system can ultimately implement the previously generated action plans through appropriate actuators and control and regulation technology. In the course of implementing autonomous navigation, the department uses approaches such as Model Predictive Control (MPC). Using tools such as Matlab and Simulink, analytical questions such as “Is the existence of a stable solution guaranteed?” and numerical questions such as “How can a solution be calculated as quickly as possible?” can be posed. The combination with the AI components is also interesting, which themselves recognise and correct errors in the system description or parameter selection.

What can TITUS Research do for partners?

With its expertise and the knowledge and experience gained from relevant projects, TITUS Research offers companies and scientific institutions comprehensive support in the development of autonomous systems, regardless of the stage of an idea. Thus, it is possible to identify potentials with the help of modelling and simulation tools to derive an approach or evaluate an existing approach with regard to its feasibility and efficiency. Based on this, a detailed analysis can be used to define requirements for the individual use case, transfer them into an implementation concept and use it to build a demonstrator. The design and establishment take place in close cooperation with the end-user so that we also provide support with technical questions and the removal of obstacles. TITUS’ research work aims to develop processes efficiently and ensure a high acceptance level with future users.


Ira Hüppe

Head of Department
Systems & Missions

+49 3375 52309 22