Robots That Learn to Use Improvised Tools

With this data, we can fit a model that proposes sequences of actions that enable the robot to use objects in the current scene as tools. We use a sampling-based planning procedure that leverages the action proposal and video prediction models and allows the robot to accomplish a variety of tasks with a number of different tools and objects. We demonstrated how our approach, which combines imitation and self-supervised interaction, can enable robots to accomplish complex multi-object tasks with a multitude of objects, and even use improvised tools under new scenarios.

Source: bair.berkeley.edu