Active Constraints (ACs) are high-level control algorithms deployed to assist a human operator in man-machine cooperative tasks [1], and define regions within which it is safe for the robot to move and cut [2]. To enhance the performance in cooperative surgical tasks, adaptive constraints have been exploited to optimally adjust the provided level of assistance according to some knowledge of the task, hardware or user. In [3] Hidden Markov Models were used for the run-time detection of the user intention to leave a guidance constraint to circumvent an obstacle. In this work, we present a novel, Neural Network (NN)-based method for the runtime classification of intentional and unintentional violations of ACs, that is trained on either statistical or frequency features from the enforced constraint forces. We investigate which set of parameters yield faster and more reliable classification results, both for guidance and regional constraints.
Recognition of Intentional Violations of Active Constraints in Cooperative Manipulation Tasks
DE MOMI, ELENA;FERRIGNO, GIANCARLO;
2015-01-01
Abstract
Active Constraints (ACs) are high-level control algorithms deployed to assist a human operator in man-machine cooperative tasks [1], and define regions within which it is safe for the robot to move and cut [2]. To enhance the performance in cooperative surgical tasks, adaptive constraints have been exploited to optimally adjust the provided level of assistance according to some knowledge of the task, hardware or user. In [3] Hidden Markov Models were used for the run-time detection of the user intention to leave a guidance constraint to circumvent an obstacle. In this work, we present a novel, Neural Network (NN)-based method for the runtime classification of intentional and unintentional violations of ACs, that is trained on either statistical or frequency features from the enforced constraint forces. We investigate which set of parameters yield faster and more reliable classification results, both for guidance and regional constraints.File | Dimensione | Formato | |
---|---|---|---|
Arico-HSMR-2015-Proceedings-FINAL.pdf
accesso aperto
:
Publisher’s version
Dimensione
229.8 kB
Formato
Adobe PDF
|
229.8 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.