The Wireless Terminal Chipset group in Lund drives innovation for the Huawei Wireless Terminal Chipset products. We design both advanced baseband architectures and receivers algorithms for both LTE and 5G chipsets.
Wireless ultra-reliable-low-latency (URLLC) applications have obtained a significant interest in academia, standardization and industry. It is essential that their performance is evaluated carefully before they are supported in the specification or eventually will be implemented in a product. For performance evaluation of digital communication systems closed form error analysis is generally preferred. However, it is not always tractable mathematically, especially for complex systems.
Then, the Monte Carlo (MC) method becomes a common choice to evaluate performance, which collects data from statistically independent simulation runs to estimate target metrics. In order to obtain reliable performance results, a rule of thumb is that about 100 errors should be collected. Due to the ultra-low error rate requirements of 10-6 or even lower conventional MC analysis takes far too long time, which is severely hindering the introduction of new features. The following example serves to illustrate the problem researchers and standardization engineers are facing: For services with more relaxed requirements (such as eMBB), the target error probability is usually 0.1 and the required number of simulation runs becomes then 1000. For URLLC simulation runs would be required.
Assuming that the simulation time for eMBB is 1.5 minutes, with the same computational power, simulation for URLLC will take 2500 hours (almost 15 weeks). To shorten the timespan at least to a level which is useable in between standardization meetings (below 1 week) the computational power needs to be increased at least by a factor of 100. Therefore, the simulation of an MC estimator is often simply unaffordable for URLLC requirements.
Objective of the Thesis:
Study Importance Sampling as a tool to significantly speed up the required simulation time for ultra-low error rates in wireless communication systems. The principle behind this approach is to bias the probability density function (pdf) to encourage more errors. By application of the so-called importance weights the calculated estimator can be kept unbiased. The applicability and achievable performance gains with importance sampling shall be investigated for un-coded and coded systems under the assumption of AWGN and fading channels.
- Master student in Electrical Engineering, Statistics, Engineering Mathematics, Engineering Physics
- A solid theoretical background in areas such as information theory and signal processing
- Basic knowledge of OFDM
- Good knowledge in Matlab or Python simulation
Start: January 2022
Dzevdan Kapetanovic, firstname.lastname@example.org