Thesis in Transfer Learning for Data-Based Control of Hydraulic Cylinders

At Bosch Corporate Research we focus on developing learning control systems for hydraulic cylinders to enhance advanced excavator assistance functions. During your thesis you will explore methods for transferring knowledge, data and model parameters between similar learning tasks to increase sample efficiency for new tasks. You will investigate measures to evaluate the performance of fine-tuned controllers to ensure they meet desired operational standards. Furthermore, you will develop strategies to reduce the computational load of model fine-tuning to enable real-time model adaptation during regular operation. Finally, you will create methods for online validation of model adaptations to ensure reliability and performance in dynamic environments.