SnailScout

An experience after the rain

I first noticed the snail during a stroll through the city after a downpour. As crowds hurried for shelter, water swiftly pooled along kerbs, yet the snail advanced slowly and resolutely between damp steps and wall bases. In that moment I realised: within a city governed by speed, efficiency and linear logic, the snail’s trajectory was entirely contrary — it thrived along moist, shaded and clingable paths.

To understand more precisely how snails interact with the micro-surface of the city, I observed the minute movements of their foot and tentacles through a magnifying glass, placing them on surfaces of varying texture, humidity and incline. A snail prioritises avoiding rough or high-resistance surfaces; when it meets a height difference it pauses or detours rather than forcing its way over, seeking alternative paths with more stable contact.

Slow movement in a fast city

Modern cities champion “rapid transit”, whereas the snail’s ecological mode of movement embodies a “slow, stationary approach”. This aligns with the natural pace of vulnerable groups — the elderly, children and pets. A systematic investigation of urban micro-environments revealed that most urban spaces adopt high-hardness, low-moisture-retention, sharply defined surfaces that prioritise efficiency and drainage, imposing invisible pressures on people who need stable support, continuous tactile surfaces and gentle transitions.

Bionic agent — path planner

What if we could translate the snail’s spatial sensing strategies into a design methodology? Imagine training a bio-inspired snail robot — not slow like a real snail, but capable of simulating its path-preference logic to identify and highlight safe, comfortable and continuous slow-movement pathways in real time.

What if urban planners, before drawing straight lines on maps, first allowed such a robot to physically “walk” the ground? In this way, vulnerable bodies no longer have to adapt to the city — the city adapts to them.

Case studies in bio-inspired algorithms — ant colony optimisation, starling flocking and the dung beetle optimiser — informed how a single slow “agent” could produce useful spatial data.

Watching snails choose

The simulation field recreates urban temperature, humidity and obstacles in order to observe and record the snails’ movement patterns. Each session has no strict time limit and usually lasts 30–60 minutes, continuing until the snails stop moving — meaning they have found a resting place. The recorded paths are converted into heatmaps and movement vectors to identify preferred routes, obstacle-avoidance behaviour and rest zones.

The study is non-invasive and focuses solely on recording movement patterns; the two snails, Andy and Betty, were cared for throughout and after the project.

Reflection: the snails had too much energy, so they tended to explore instead of settling. Is the experimental field too small for them?

Field 2 gave the snails a bigger space outdoors with randomly placed walls and more distributed heating and cooling areas. Field 3 added high coverage, a clear distinction between large and small obstacles, and a wider temperature range.

Reading the trails with DeepLabCut

DeepLabCut is an open-source, deep-learning-based toolkit for markerless pose estimation. I used it to track keypoints on the snails across every recorded frame and turn their movement into coordinate data. The snail heads were mostly concentrated between coordinates 700–1000, which led me back to the labelled images to investigate further.

Observations suggest that snails do not exhibit absolute avoidance of unsuitable environments, but rather adjust their behaviour based on a prioritisation of the conditions they encounter.

• Temperature firstPrioritise the optimal temperature of 20–28 °C.
• Then humidityPrioritise optimal humidity of 75–95% while avoiding barriers.
• OverrideIf the temperature exceeds 30 °C, switch to the alternate pathway despite existing barriers.

From snail tags to a snail robot

To capture the microclimatic relationships between different points in the city, I first explored tracking real snails. Harmonic radar uses a passive tag to reflect incoming radar signals at a higher harmonic frequency; the tags are extremely light, which makes them ideal for small organisms. Because traditional tags are glued on and cannot be reused, I tested a series of biomaterial patches — waterproof on top, gently adhesive and easy to peel on the bottom.

The radar proved fragile, with a limited and unstable detection range, so I moved on to a disassembled mini GPS. That was still too large for a snail to carry, and consumer trackers update only at discrete intervals. This led to the turning point of the project:

What if the tracking subject is not an actual snail, but a device that simulates snail detection and behaviour?

Building a body that crawls

The robot needs to detect (a soft, deformable body with temperature, humidity and camera sensing), respond (avoid temperatures above 30 °C, prefer 70–90% humidity, avoid barriers unless both other conditions are unmet) and record (a location-preference diagram via GPS).

From research into soft robotics and snail locomotion — muscular pedal waves move the head first, then the shell — I cast pneumatic bodies in Eco-Flex silicone. The first prototype barely changed volume. Focusing on front expansion produced about 5 mm of forward movement per cycle; iterating the “foot” section with nano tape for friction eventually gave a walking length of around 2 cm with the supporting friction the robot needed.

Sensing, powered by the sun

To keep the robot small I chose the smallest possible boards: an ESP8266 and Nano, a GPS module, a temperature-and-humidity sensor and a camera, charged by a flexible solar panel through a charge controller. Testing showed the camera could replace ultrasonic sensing — by processing images it can identify obstacle location and shape — so the ultrasonic module was removed to reduce pressure on the soft body.

The shell and soft-body moulds were 3D printed, the body cast in silicone, and the components assembled with care for the snail shell’s centre of gravity.

SnailScout in the city

SnailScout determines its path based on temperature and humidity inputs, following the preference order observed in the snails. In the current prototype the logic is simplified: the air pump responds when conditions fall outside a test range of 40–60% humidity or 20–25 °C, steering the robot away from those areas. At every movement it uses the camera and GPS to locate itself and annotate the collected data onto the corresponding positions.

• Snail shell lidLets users check the components and export data easily.
• Preference hardwareModules that imitate the snail’s spatial preferences.
• Solar panelPower input.
• External air pumpProvides locomotion power for the soft robot.
• Soft robot bottomEmulates the snail’s gradual motion for high-resolution documentation of its immediate environment.

Limitation: achieving fully functional operation will require further programming, and a strategy for controlling the volume of the associated air pump is yet to be determined.

Empathetic design for multi-species futures

Urban planners can use micro-scale observations of temperature and humidity to locate comfortable zones (20–25 °C) for public spaces and high-density development, and mitigate unsuitable zones with cooling strategies, green infrastructure and density adjustments. Weekly or monthly reports could help planners monitor and understand otherwise uncontrollable extreme weather events.

This project highlights the potential of integrating non-human perspectives into design — envisioning urban environments and interactive systems that are more inclusive, adaptive and responsive to the shared preferences of both human and non-human species.