Abstract New meta-materials are developed with the usage of piezoelectric transducers’ networks. Within the number of controlling strategies for vibration mitigation, this study uses the classical derivative control law as a basis. As a preliminary study in optimization with AI, an automatic algorithm using Reinforcement Learning (RL) approached with Trust Region Policy Optimization (TRPO) tunes a controller on an experimental cantilever beam. The control law is a simple derivative feedback between two collocated piezoelectric transducers close to the beam-clamped end. The RL algorithm runs offline on an estimated model of the experimental setup. The study compares control tuning methods between Reinforcement Learning results and a classical published approach.