The Neuromorphic Computing Solutions Group aims to develop an innovative neuromorphic toolset that enables the seamless integration of neuromorphic technologies into existing systems. By improving the interoperability of software and hardware, conducting targeted research on optimised training algorithms for spiking neural networks (SNNs), and employing a benchmarking-based system design approach, the group enhances the efficiency and effectiveness of these technologies.

Using proven methods from the fields of data fusion, control systems, and industrial software development, the Neuromorphic Computing Solutions Group is tapping into the market for neuromorphic COTS sensors and processors. The focus is on developing robust and resilient end-to-end solutions and promoting industrial adoption through training programmes.

The group’s research approach is based on Prof. Axenie’s practical expertise in the design, development, and implementation of closed-loop control systems, particularly in the fields of industry, robotics, communication systems, and mobility. By overcoming existing obstacles, the Neuromorphic Computing Solutions Group makes a decisive contribution to the seamless, cross-industry adoption of neuromorphic technologies.

Objectives

The novel toolset encompasses efficient training, optimisation, and flexible porting of neuromorphic neural networks to improve integration into existing architectures and promote the interoperability of software and hardware. It forms the basis for the implementation of reference designs and benchmarking and takes into account aspects of data fusion and heterogeneous architectures combining conventional and spiking neural networks.

The system design follows a benchmarking-based approach. In this process, benchmarking is performed at the system level using concrete, industry-relevant use cases. Automatically generated reports serve as the basis for customer interaction, for example, to quickly select target hardware.

Based on the implemented reference designs, application-specific demonstrators are developed. The demonstrators facilitate industry interaction and build confidence in the technology.