Vision-based deep execution monitoring
This work addresses the challenge of reliable robot task execution in unknown settings, offering a domain-specific solution that is incremental by building on existing deep learning and reinforcement learning techniques.
The paper tackles the problem of improving robot action execution monitoring in uncharted environments by using visual perception to verify preconditions and postconditions, and it achieves this by integrating DCNNs for object recognition, non-parametric Bayes for relation discovery, and deep reinforcement learning for visual search policies to recover from failures.
Execution monitor of high-level robot actions can be effectively improved by visual monitoring the state of the world in terms of preconditions and postconditions that hold before and after the execution of an action. Furthermore a policy for searching where to look at, either for verifying the relations that specify the pre and postconditions or to refocus in case of a failure, can tremendously improve the robot execution in an uncharted environment. It is now possible to strongly rely on visual perception in order to make the assumption that the environment is observable, by the amazing results of deep learning. In this work we present visual execution monitoring for a robot executing tasks in an uncharted Lab environment. The execution monitor interacts with the environment via a visual stream that uses two DCNN for recognizing the objects the robot has to deal with and manipulate, and a non-parametric Bayes estimation to discover the relations out of the DCNN features. To recover from lack of focus and failures due to missed objects we resort to visual search policies via deep reinforcement learning.