Reinforcement Learning Projects
My research project in Robot Learning Lab.
I studied numerous papers on various domains of Reinforcement Learning such as meta-learning, multi-task learning, and model-based learning in the first semester to identify a topic for further study.
In the second semester, I focused on the ‘Multi-Task Decision Transformer’, to realize skill merging through offline learning. My work involved generating datasets, defining reward functions, and masking to train the model. I successfully merged tasks like “rotation”, “reaching the goal”, and “avoiding lava” in the Gym’s Minigrid environment.
