Autonomous Reply develops embedded software in C/C++ for autonomous systems, with a special focus on real-time capability, maximum performance and optimal memory utilisation. For this purpose, various application frameworks and platforms are used. CODESYS is an example of an application implementation in an industrial environment, and AUTOSAR in the automotive sector.

For special subject areas such as navigation localisation solutions and deep learning approaches, the specialists use ROS/ROS2 and NVIDIA ISAAC. However, for this a powerful edge device is required.

Model-based function control
Autonomous Reply develops software and functions tailored to needs, requirements and available budget. With model-based function development, the developed software components can be tested on prototype ECUs at an early stage of development, which minimises time-to-market. Using automatic code generators such as TargetLink, the models developed in Matlab, Simulink and Stateflow can be generated as production code. This allows early verification through simulation and testing.

In addition, Autonomous Reply supports both the development-accompanying and the final virtual testing of the functions and algorithms. The portfolio ranges from initial model checks to MiL and SiL tests to embedding the software in virtual environments such as CarMaker with subsequent simulations and their evaluations.

In summary, the focus for rule algorithms is on the following topics:

• Localisation (Indoor & Outdoor)
• Longitudinal and transverse guidance
• Path and trajectory planning
• System status functions

Autonomous Reply develops innovative software solutions for autonomous systems with a focus on edge computing and software-defined systems in Python, Java C/C++. Especially in the areas of deep learning, navigation and localisation, where highly complex algorithms and powerful controllers with maximum performance are required, the experts rely on software frameworks such as ROS/ROS2 and NVIDIA ISAAC.

Another focus in this topic area is cloud application development, which is characterised by cloud-based software development and a direct connection to the edge device. By means of code creation and a CI/CD functionality for direct deployment on the edge device, the autonomous vehicle can be kept up-to-date via software updates. This creates a flexible balance between applications on the edge device and applications in the cloud, enabling high-performance hybrid solutions for HD Maps, navigation and fleet management, to name a few.

ROS applications
Autonomous Reply offers knowledge and expertise for the conception, design, implementation and support of customised solutions in the areas of perception, navigation, control and simulation.

NVIDIA ISAAC applications
In the field of NVIDIA's ISAAC robotics platform, Autonomous Reply provides all the steps in the development of a software application for autonomous things. This includes the integration of existing ISAAC components, for example for localisation and mapping (SLAM), trajectory planning or obstacle avoidance, as well as the development of new ISAAC components. Beyond this, the experts create new ISAAC sim scenarios and use them to realistically test use cases or to generate training data for Deep Learning based computer vision algorithms (for object classification, 2D/3D object detection or semantic segmentation).

The most important elements of autonomously moving things are object recognition and classification capabilities as well as tracking and prediction of the future pose/position of objects.

Autonomous Reply therefore offers a comprehensive service especially in the areas of computer vision and deep learning. This includes data identification and selection for a representative dataset, identification of a suitable architecture for and training of the neural network as well as optimisation of the models with regard to accuracy and inference times. Last but not least, Autonomous Reply also takes care of the integration of the elements on the target platform.

The experts develop and train efficient deep learning models for this implementation. In doing so, Autonomous Reply uses both native cloud services and its own neural networks. In addition, the experts also offer services in the field of deep learning, which are used for navigation, trajectories/path planning and motion prediction.

Autonomous Reply helps to concretise the challenges around the topic of Autonomous Things and to define the appropriate solution paths in order to achieve goals efficiently and precisely.

The team advises on the best possible end-to-end strategies for autonomous systems, starting with advice on system architecture, the selection of hardware and the appropriate communication channels. In a cross-functional team, Autonomous Reply develops the best possible solution for each application. The advisory services include:

• Strategy Development
• System Architecture Design
• Process Advisory
• Automotive Project Management
• Project setup and Supervising
• Exploration of HW & Sensorics for Autonomous Vehicles