Tooling for the deployment of AI-based scheduler on the Xilinx Versal

Thesistype: Bachelor-/Master-/Student Thesis
Contact: Interested students in this area should please contact Dr. Daniel Onwuchekwa (
daniel.onwuchekwa@uni-siegen.de, H-E 010)


Description: A scheduler decides which task is to be executed, the time of the execution, and which processor will execute a task in cases where we have more than one processor. These three decisions can be carried out before the program starts(design time) or during its execution (run time). An AI-based scheduler is designed to support fast adaptation times to different context events during run-time. The AI-based scheduler is realised by learning a given scheduler using machine learning techniques. The learned model is then deployed to generate new schedules at run-time using inference. It is therefore, required to design a tooling for the Xilinx Versal Platform to deploy the learned model for a state-of-Art time-triggered Network-on-Chip. The Xilinx Versal is a fully software-programmable, heterogeneous compute platform that combines Scalar Engines, Adaptable Engines, and Intelligent Engines. It combines three architectures, scalar processing elements (e.g., CPUs), vector processing elements (e.g., DSPs, GPUs), and Programmable Logic (e.g., FPGAs). It is intended to deploy and utlise the learned model in the programmable logic.

Task Objectives
You are required to model and design the tooling to deploy and configure the AI-based model.

Desired Skills


Language
English

Seminar topic in the field of SWARM Intelligence

Thesistype: Student papers
Contact: Interested students in this area should please contact Daniel Onwuchekwa (
daniel.onwuchekwa@uni-siegen.de, H-E 004)


Description: Swarm intelligence (SI), which is an artificial intelligence (AI) discipline, is concerned with the design of intelligent multi-agent systems by taking inspiration from the collective behaviour of social insect colonies and other animal societies. A swarm engineering approach incorporates a large number of relatively simple robots given simple commands whose interactions cause global emergent behaviour. SI can be applied to cellular robotic systems, which consists of collections of autonomous, non-synchronized robots cooperating on a finite n-dimensional cellular space under distributed control. SI is also predominantly used for optimization tasks and has great potential in handling large size complex optimization problems and capable of producing results in very less time.

Task Objectives
Explore swarm intelligence algorithms including simulation tools

Language
English