Touch sensing on a desktop pet: where the zones go and what they feel like

Desktop pets earn their keep through touch, and the entire experience is decided by where the sensing zones sit and how the firmware interprets them. Getting this right is part industrial design, part behavioural engineering.

Map the petting, not the enclosure

Owners pet the head, stroke the back, and fidget with whatever their hand rests on. Zones that follow those instincts — a crown zone, two flank zones, a chin tickle spot — feel designed; zones that follow convenient PCB real estate feel random. We prototype zones in the physical dummy with conductive tape before committing the PCB layout.

Differentiation is the hard part

A single large zone reports “touched”; a product needs “stroked left”, “held”, “tapped” and the difference between a pat and a grab. That comes from the electrode geometry (a gradient of two electrodes reads direction), from sampling rates, and from a gesture classifier in firmware. The classifier is where cheap products cut corners — and where users notice.

False triggers have worse consequences than dead zones

A zone that fires when the robot is carried, when another device touches it, or when a hand passes nearby trains the owner to distrust it. Guarding — shield electrodes, debouncing, carrying-mode logic — costs board area nobody budgets for up front. We budget it in every proposal.

Touch needs a conversation with motion

The pet should lean into a pat. That means the behaviour engine reads touch and motion together, and the mechanism responds while the hand is still there. Touch and motion designed separately produce a robot that pets you back a second too late.