Learning to Detect Spatial Regions with Constraints

This paper proposes a method for constraint detection using only proprioceptive data. This method for constraint detection was tested on a pre-existing data set and was found to be able to detect insertions. Two additional experiments were performed. The first showed that the constraint detection algorithm can be used to associate an action with an event. The second experiment implemented a learning technique to show that learning from constraint detection can improve the efficiency of performing a task. Our method can be applied to both environmental constraints (i.e. limited space to move) and object constraints (i.e. levers, wheels, etc.).