Short name: DeepRL4MEC
Long name: Deep Reinforcement Learning for MPEG DASH segments’ dynamic assignment in MEC environments
Call Stage 1: F4Fp-SME-COD1 (see call details)
Proposal number: F4Fp-SME-COD200428-04
SUMMARY REMARKS & TESTBEDS
The computing resources within Mobile Edge Computing (MEC) application servers are limited. For example, a given server might be CPU-stressed while servicing many clients. Within this call, we shall extend our prototype to further improve clients’ performance by implementing and validating two new features: (I)we will rewrite parts of our code in order to undertake mapping decisions per each client MPEG DASH segment, not for the whole MPEG DASH file. This way, we will be decided at a much faster scale than our current implementation, following more closely the changes that occur in the MEC environment; and,(II) to lower the training time of our models, we shall employ GPUs.
- DeepRL4MEC experiment – Stage 1 | FEC9 (download the poster)
- Review DeepRL4MEC experiment – Stage 1 | FEC9 (download the slides)