Current students


Section: Computer Science and Engineering

Major Research topic:
Advances in Benchmarking of Robot Tasks and Systems

Benchmarking components of autonomous robots requires taking into consideration many aspects since their performance depends on the robot’s sensors, the quality of the information from other components, and the characteristics of the environment, to name a few. In most cases, the benchmarking methodology in autonomous robotics neglects these dependencies and focuses on subsystems without considering the systems they are going to be part of.
The objective of this thesis is to develop a methodology to evaluate the performance of a software component with regard to the characteristics of the system and the environment. To obtain such a comprehensive description of a component's performance, we believe a statistical model (that we call component performance model) conditioned on several components is required. The performance can thus be estimated from features of the system and environment that we set up or measure. We call this procedure performance modelling to differentiate it from plain benchmarking of single isolated independent components or systems.
A further objective of this thesis is to study the composition of performance models of different components and estimate the performance of an entire system at design time. This is indeed not possible nowadays, and complex robotics systems are usually the result of a long and tedious trial and error procedure. By enabling developers to predict the performance of a system beforehand, we aim at reducing the trial and error process to just a few iterations.