STAT.AI has participated with our presentation called Advanced Control Strategies Based on Reinforcement Learning for Linear Actuators as part of the The 2nd International Electronic Conference on Actuator Technology.

The evolution in linear actuator control demands solutions that go beyond classical methods, such as PID control, to address challenges related to nonlinearities and uncertain conditions. In this work, the use of reinforcement learning (RL) algorithms is explored to develop an advanced controller applied to a second-order dynamic model. Using Python libraries, a simulated environment was designed where a Policy Gradient (PPO) RL agent optimizes its control strategy through continuous interaction and adaptive rewards.

The results of the RL controller were compared with a PID system, showing that the RL-based controller outperformed the PID in accuracy. Although training the RL agent involves greater computational complexity, the results demonstrate its potential as a robust alternative for complex dynamic systems and changing conditions.