Computer Vision News - September 2021

progress, we propose to train the task generator by balancing the robot's performance in the generated tasks and the similarity to the target tasks. Through adversarial training, the task similarity is adaptively estimated by a task discriminator defined on the robot's experiences, allowing the generated tasks to approximate target tasks of unknown parameterization or outside of the predefined task space. 17 Kuan Fang Learning Sequential Tasks from Task-Agnostic Data. We present a method that learns to solve sequential tasks by learning from task-agnostic interaction data . To effectively find plausible plans, we propose to hierarchically generate subgoals and actions using learned latent representations. We propose cascaded variational inference to learn the latent representations fromtask-agnostic interactions between the robot and the environment. We demonstrate that our method successfully solves sequential manipulation tasks under various semantic and physical constraints in cluttered environments . Generating Tasks for Curriculum Learning. We present a framework to progressively generate tasks as the curricula to facilitate reinforcement learning in hard-exploration problems . At the heart of our approach, a task generator learns to create tasks from a parameterized task space via a black-box procedural generation module. To enable curriculum learning in the absence of a direct indicator of learning