SHMUC Use Case  D1

Making Tea With Sugar: (Ver. 1.0)


SHMUC prototype  >  D_Behaviour_Pattern_Abnormalities_Use_Cases  >  D1_Making_Tea_with_Sugar



Mary wishes to make a cup of tea

Actors Mary, the Smart Home, Debbie

Initial State

Mary was at home alone.


During the training phase, the system identified a tea-making behaviour that consisted of the following actions (where [.]0,1 describes an activity that occurs 0 or 1 times):

[go to sink-bench]0,1
[turn on tap
, fill kettle, turn off tap]0,1
[boil water]1
[put tea in pot]
[pour water from kettle into pot]
[go to fridge, open fridge, take out milk, close fridge, go to sink-bench]
[pour milk into cup]0,1
[pour tea into cup]
[spoon sugar into cup]

After training, this system was ready to monitor Mary’s activity. One afternoon Mary made a cup of tea without milk, but with sugar, using the following set of actions:

go to sink-bench
turn on tap, fill kettle, turn off tap
boil water
put tea in pot
pour water from kettle into pot
spoon sugar into cup
pour tea into cup


This sequence does not match the learnt pattern exactly, as the items in bold are reversed in order with respect to the syntax created during training, and the sequence was considered as a novelty by the system. However, it did not cause an immediate alarm, as the system identified that the order of two actions, i.e. “get sugar” and “get hot water”, does not affect on the final state of the activity. Therefore, the system did not create an alert, but modified its representation of tea making instead.

Norm The tea-making sequence conforms to the syntax specified by the Finite State Machine


The activity pattern was automatically updated, and a notification was issued.

System Design Implications

This use case deals with multiple valid activity orderings.

Activities often comprise a partially ordered sequence, and there is no guarantee that  observation of any number of instances will reveal all the orderings. It is therefore important that the system should incorporate a mechanism to recognise that the events that make up an activity have occurred out of the normal sequence - or equivalently, that an order of events that resembles, but does not correspond exactly to any previously observed order may be a previously unseen but valid order.

The system could not infer this without external input from a competent source (which might rule out the inhabitant, if the inhabitant were dementing)