Environment detection and motion planning for autonomous systems

An essential component of autonomous systems are methods for autonomous environment perception, position estimation as well as static and dynamic path planning. For autonomous environment detection, the data of different sensors have to be processed and transferred into environment models. AI methods help here, for example, to derive driveability maps from camera images and laser data. We have at our disposal powerful methods of sensor data fusion and Simulataneous Localization and Mapping (SLAM) to enable task- and environment-specific optimal environmental detection and navigation. This basis is used by our motion planning algorithms to optimally control robot systems in a wide variety of environments (structured, unstructured, underwater).

Fields of application

The Fraunhofer IOSB coordinates the Competence Center ROBDEKON ("Robots for decontamination in hostile environments"), which was founded in 2018. Here, complex autonomous systems for use in hostile environments are developed and demonstrated. Comprehensive solutions for autonomous movement and manipulation are demonstrated e.g. on autonomous construction machines.

In the project AKIT (“Autonomy KIT for near-series working vehicles for networked and assisted recovery of hazardous sources”, text in German language only), a kit for work machines was developed, with which excavators and tractors that are available on site in disaster scenarios can be quickly converted into autonomously and semi-autonomously operating recovery and transport vehicles to carry out work in particularly hazardous areas. This successfully demonstrated concept has been under further development in the AKITpro project since 2021.

In LocSens(text in German language only), the institute is involved in the development of multi-sensor systems for environment detection under harsh conditions. In areas where optical sensors cannot be used, radar and UWB-based sensors help to reliably detect the environment. The solutions implemented here can be transferred to many new use cases for autonomous systems.

The solutions and algorithms developed for the various applications can be quickly transferred to customer-specific applications due to their modularity. In many cases, ROS (Robot Operation System) acts as middleware. Partial solutions can also be offered as embedded implementations on powerful microcontroller platforms.