Case Study: Cobot Case

The interesting events from CoBot's standpoint are those, which were not expected or least expected to happen, from the CoBot's past task execution data. We define these interesting events as anomalies-- deviation from the expected data. The expected value for an event can be computed from the respective log table, we create by analyzing the bag files. Using the expected data we identify the instances which are anomalies, and verbalize them comparing it with a past instance or the expected data for that event.

Here the procedure has been discussed in the context of describing interesting events considering the time taken by the CoBot while executing a task. We find anomalies in time taken to cover segments of the path plan belonging to different
We then give an anomaly score $A_i$, for the segment $i$ in the path plan which lies in the corridor CID. The anomaly score is calculated using the following equation:
\\ \\
$A_i = \frac{k1}{f_{CID}(V_{t,CID})}$, $\forall$ i = 1... N, where k1 is a constant.
\\ \\
Using the above procedure all the segments in the path plan of the current task are assigned an anomaly score. The verbalization algorithm then considers segment i to be interesting if it is greater that a threshold $A_{th}$.
\\ \\
$A_i > A_{th}$, $\forall$ i = 1... N
\\ \\
For each segment $i$, that are considered interesting, we then find the expected time $T_{e,i}$, that CoBot take usually to travel such a segment, and the time that CoBot took $T_{t,i}$, for covering the segment $i$ in the corridor, which is
\\ \\
$T_{e,i} = \frac{|| Start\ point_i - Ending\ point_i ||}{\argmax\limits_{V} \ f_{CID}(V)}$ \\
$T_{t,i} = End\ timestamp_{t,CID} - Start\ timestamp_{t,CID}$
\\ \\
We then tag the segment $i$ with the following utterance comparing the two times.
\\ \\
{\it `Usually it takes me $T_{e,i}$ seconds but this time it took me $T_{t,i}$ seconds to travel in corridor