Galileo enhancing micromobility performance: the experience from Reby

Improved accuracy and continuity, better security, among others: the benefits of Galileo for mobility plans and solutions might be known, but what are some of the concrete use cases of actual mobility players leveraging Galileo? Introducing Reby: a Spanish company that designs, manufactures and distributes shared electric vehicles, using Galileo technology built in their solutions.

Reby offers a practical and sustainable solution to move around cities. Their team is composed of more than 100 people and have more than 306.500 registered users. Currently, their fleet including their own vehicles and the ones produced for their franchisees consists of more than 6.000 kick-scooters, 2.000 bicycles and 3.000 mopeds. Reby’s users have travelled thousands of km, hereby saving many tons of CO₂, and helping to keep cities clean and sustainable.

Let’s hear their story about how and why they are using Galileo for their services.

Why invest in Galileo

In the micromobility business, it is paramount to know the position of a vehicle at all times. Therefore, all Reby kick scooters, mopeds and bikes are equipped with GNSS receivers which allow to monitor their position at all times. This enables users to quickly find a vehicle they have booked, the operations team to work more efficiently, the data team to optimise routes, determine demand clusters, and in general perform geospatial analysis. Furthermore, it enables to comply with local regulations as well as detecting suspicious behaviours such as a vehicle being stolen or a user travelling outside the limits of the service area.

Problems and challenges

An error of up to 50 meters can be ignored when inspecting a city-wide map or a continental map, for example. Unfortunately, the error of the measurements of a standard GNSS receiver cannot be ignored at street level resolution. This is a classic problem in GPS and GLONASS which the European GNSS constellation Galileo aspires to solve by reducing the error component of the measures up to a few centimetres.

The source of the error is mostly due to multipath fading of the received GNSS signal. This is especially critical in urban environments, where the GNSS signal is likely to be received multiple times with different power levels and time delays.

Imprecise GPS hurts the business

The likelihood that a user will use one of Reby’s vehicles is strongly related to how easily they can find them. If they cannot find the vehicle where it is expected to be, they will look for an alternative. In addition to this, they will consider this as a bad experience which will lower the acceptance and level of satisfaction with the service. Moreover, many users use multiple micromobility services. While looking for a Reby vehicle, they are likely to come across vehicles from the competition in high demand areas and the probability that they use Reby’s service will decrease as well. This is an issue which undoubtedly reduces revenue.

As a data-driven company, a great number of Reby’s daily decisions are based on their own data as well as external sources. Precise data helps increase the confidence of people involved in decision making.

Operationally, it is also painful to have bad positioning signals. The logistics and maintenance cycle are very time consuming and the staff responsible for deployment, battery-swapping and maintenance can lose a great deal of time looking for these vehicles if they cannot be located easily, incurring additional costs which over time amount to considerable expenses.

More restrictions require better data

The days when companies would place kick scooters anywhere on sidewalks and bike lanes are long gone. Countries and City Councils demand regulation, which Reby have always welcomed and embraced as a means to sustain and fully develop the industry and their business. More often than not, a subset of these regulations is spatial. From designated parking areas to speed limits in certain roads or areas, all of them are weak around the edges. Setting up geofences for parking areas of 10 m² or 20 m² is futile considering that the GPS error can be of up to 50 meters in urban areas. And even if the geofence was much bigger, it would still be problematic close to the edges. The same issue arises when trying to set speed limits to individual streets or bike lanes. The location error is much bigger than the width of the street, thus rendering the solution useless.

With centimetre accuracy, this system could be trusted, but not with the current state-of-the-art of GPS technology alone.

We are prepared

Using more accurate GNSS like Galileo can increase the accuracy of the vehicle position and thus the quality of the service. We must ask ourselves if we are prepared to implement it. And the answer is YES. Reby’s IoT hardware installed in their vehicles is capable of changing the position receiver between different satellites, i.e., GPS, GLONASS or Galileo.

Other solutions

So far, we know that the actual GPS locations are often not accurate enough, but we can still implement different data processing solutions to improve accuracy. We describe two different approaches of data pre/post-processing.

Kalman/Particle Filters

In control theory and statistics, the Kalman Filter is an algorithm that uses a series of measurements observed over time, taking into account statistical noise and other inaccuracies, and produces estimates of unknown variables that tend to be more accurate than those based on a single measurement alone, by estimating a joint probability distribution over the variables for each timeframe.

Kalman Filter would help us by predicting the next position (k+1) during a ride, considering the previous position (k) and a defined model. However, these new positions would only be predictions so we cannot be sure to replace them from the original GPS positions.

The issues that come with this solution are related to real-time processing, which is costly and sometimes unavailable until later, and the added vehicle-sensorisation required to calculate the state-transition matrix – compass/gyroscope.

Map Matching

The most common map matching approach is to take serial location points and relate them to edges in an existing street graph (network), usually in a sorted list representing the trip of a user or vehicle.

Map matching algorithms can be divided in real-time and offline algorithms. Real-time algorithms associate the position during the recording process to the road network. Offline algorithms are used after the data is recorded and are then matched to the road network. Real time rectification is computationally more expensive than doing it just one time at the end of the ride, and the results are pretty much the same.

Reby’s R&D team have recently participated in the Moliere project Hackathon 2021 presenting a hybrid solution to solve the challenge of inaccurate geopositioning, which includes map matching (plus clustering of similar points) and rectification of final ride positions to moped parking spots located in the city of Barcelona. This solution was awarded with the first prize in the Molière hackathon.

One of the big challenges that we face with this solution has to do with the street network available for matching. Certain assumptions can be made about cars or mopeds that cannot be made of other vehicle types such as e scooters, bicycles or similar. For example, the types of roads a car can drive on help in limiting the available network. Also, almost all road-available streets are mapped somewhere (e.g., Google Maps or Open Street Maps). This is not the case with micromobility, where driving patterns are less predictable and they do not tend to follow conventional networks (crossing a park or using pedestrian lanes).

Conclusions

Location accuracy plays a significant role in the mainstream adoption of micromobility services. The companies involved in this sector face a series of challenges which nowadays require smart and efficient solutions. That is why Reby are implementing these hardware and software solutions which help them reduce the gap between the real and tracked positions to, ultimately, give a better service to the final users and other integrated mobility platforms.