The six-month e-scooter pilot conducted in Helsinki has shown how cities can identify risk before it becomes an accident, using sensor-based AI insight data to reveal where riders struggle most – particularly around complex infrastructure and in wet conditions.
Traditional safety reports rely on accident statistics, which often miss the vast majority of minor incidents and rider challenges. As part of the EU-funded ELABORATOR project the Safety Sense Helsinki pilot bridged this gap by equipping shared e-scooters with See.Sense AI sensors, recording up to 800 data points per second. This system distinguishes between normal riding and rapid evasive manoeuvres to pinpoint where risky situations are happening.
The pilot deployed 40 sensor-equipped e-scooters, generating about 2,500 trips, covering 4,500 km, and showing a good coverage across Helsinki city center and some closeby districts. Data collection lasted for about 3 months.
The collaborative project was led by Forum Virium Helsinki, the City of Helsinki innovation company, in partnership with micromobility operator Tier-Dott, technology and data company See.Sense, and AI-powered mobility data platform Vianova.
“This completely novel dataset provides evidence where e-scooter riders are constantly forced into harsh braking, swerving, or where road surfaces cause instability in riding. This information, in turn, can be used proactively in planning of street design or guidance.” says Noora Reittu, Senior Project Manager from Forum Virium Helsinki’s ELABORATOR project.
Why This Matters for Vision Zero Cities
Helsinki is committed to Vision Zero, aiming to proactively identify and reduce risk. Integrating this novel See.Sense sensor data with Vianova’s visualisation and analytics platform demonstrates a practical tool for planners to prioritise interventions and evaluate whether changes on-street are working. Over time, this kind of dataset can establish a baseline for monitoring how infrastructure updates affect rider behaviour — enabling continuous, evidence-led safety improvement.
7 Key Findings from the Pilot
The pilot delivered a novel data set that highlights the impact of infrastructure and the unintended consequences of current digital restrictions.
1. Cost-effective method for data collection
Using shared vehicle fleets as “city-wide sensing network, delivering cost-effective insights that can be used for transport planning and infrastructure improvements.
2. Risk versus exposure -analysis revealed new risky locations
The risk versus exposure analysis, conducted on the Vianova platform, also revealed new locations within the city where the number of rapid evasive manoeuvers is high relative to the number of observations.
3. Infrastructure matters
The data successfully identified specific locations where e-scooter riders struggle. Clusters of minor incidents were consistently found at complex junctions, areas with pedestrian crossings, underpasses, tram tracks and cobblestones. This confirms that these elements significantly compromise e-scooter rider safety and control.
4. Peak commute times are the riskiest
Risk is highest during peak commute periods, with the morning peak more acute than the evening both in timing and risk. This mirrors patterns seen in cycling: morning trips are often more time-sensitive, with greater pressure to arrive at work on time, which can lead to faster riding, tighter gaps, and more interactions with traffic in a shorter window.
5. Wet weather increases risks
Wet weather increased the risk of incidents, with harsh braking events increasing by 1.7 times and swerving events by over 2 times. The locations where these events occurred during wet conditions often differed from the usual high-risk areas.
6. Geofencing should consider the surrounding infrastructure
Shared e-scooter users use the same bike lanes and roads as other vehicles, but have less control over their vehicle due to the geofence based restrictions. Categorical implementation on restriction zones based on their type (e.g parks, kindergartens) without considering the surrounding infrastructure may lead to unpredictable vehicle slowing type of behaviour increasing risks for the user.
7. GPS inaccuracies should be considered in regulation design
Categorical zoning may also have an impact on busy adjacent cycle lanes due to GPS inaccuracies. For example, a slow-speed zone near a park entrance inadvertently affected riders on a nearby bike lane.