Content uploaded by Frauke Behrendt
Author content
All content in this area was uploaded by Frauke Behrendt on Aug 25, 2015
Content may be subject to copyright.
Smart e-bike monitoring system: real-time
open source and open hardware GPS
assistance and sensor data for
electrically-assisted bicycles
ISSN 1751-956X
Received on 7th October 2014
Revised on 8th May 2015
Accepted on 12th June 2015
doi: 10.1049/iet-its.2014.0251
www.ietdl.org
Chris Kiefer1, Frauke Behrendt2✉
1
Department of Informatics, University of Sussex, Brighton BN1 9QJ, UK
2
School of Arts, Design and Media, University of Brighton, Brighton BN2 4GJ, UK
✉E-mail: f.behrendt@brighton.ac.uk
Abstract: The smart e-bike monitoring system (SEMS) is a platform for the real-time acquisition of usage data from
electrically-assisted bikes (also called pedelecs or e-bikes). It is autonomous (runs off the bike battery), replicable (open
source and open hardware), scalable (different fleet sizes) and modular (sensors can be added), so it can be used for
further research and development. The system monitors location (global positioning system), rider control data (level of
assistance) and other custom sensor input in real time. The SEMS data feeds an online interface for data analysis, for
riders to view their own data and for sharing on social media. The basic system can be replicated by other researchers
and can be extended with modules to explore various issues in e-bike research. The source code and hardware design
are publicly available, under the General Public License, for non-commercial use. SEMS was implemented on 30 bikes
and collected data during 10 months of real-word trials in the UK. This study details the design and implementation of
the hardware and software, discusses the system use and explores features for future design iterations. The SEMS
turns singular e-bikes into a networked fleet and is an example of the internet of things in the cycling context.
1 Introduction
E-bikes are equipped with a power source and therefore offer unique
opportunities to monitor and understand usage and their interaction
with the urban environment with potential benefits for both
e-cyclists and conventional cyclists. A detailed understanding of
the usage of e-bikes in specific geographical and cultural contexts
can help to understand and communicate their potential benefits
for sustainable transport and beyond.
The term ‘e-bike’in this paper refers to bicycles equipped with a
small motor and battery where riders always have to pedal, but can
switch on electric assistance (usually with a choice of low,
medium or high settings) if they wish. The assistance cuts out
when pedalling ceases or the speed of 15 m/h (25 km/h) is
exceeded. These e-bikes are often referred to as pedelecs and are
popular in many European counties. Other types of e-bikes exist,
for example, those where assistance can be used without pedalling
–these are especially popular in many Asian countries (see e.g.
[1]) but are not within the remit of this research. Many
configurations of motor and battery are possible on e-bikes [2, p5]
with the models used on this study representing two of the most
popular designs: (i) crank-driven motor with central battery and
(ii) a front-hub motor with a rack-mounted battery (see Fig. 1).
E-bikes are rapidly becoming mainstream in European countries
with developed cycling cultures, appealing to both existing and
new cyclists [3]. For example, in the Netherlands, sales of e-bikes
equal or exceed those of conventional bikes in value; in Germany,
one in ten bikes sold is an e-bike; and there are estimated to be
over a million e-bikes in use across Europe [3]. However, they are
still not well-known in England. A better understanding of how
people in the UK engage with e-bikes could help to identify issues
for policy, design and research that could lead to a higher uptake
of e-bikes. The 2011–2014 ‘Smart e-bikes’research project [4]
works on this and the smart e-bike monitoring system (SEMS) has
been developed as part of this work. The monitoring system is
implemented on a fleet of 35 e-bikes in Brighton (UK) (see Fig. 2).
This paper discusses related work, outlines the design framework
for the monitoring system, before presenting, in detail, the workings
of the hardware and software. It then shows how the monitoring
system has been used successfully by 93 participants, including
four 8-week public user trials with 20 participants each, and
explores features for future design iterations. The system has been
designed for and tested with two types of e-bikes, and could be
adapted to work with other models. The source code and hardware
design are publicly available [5] under General Public License, for
non-commercial use.
2 Related work
The tracking of moving objects requires battery power, and therefore
this has been largely implemented on those moving objects that
already have a battery –typically vehicles with motors such as
cars, lorries, boats or trains. Conventional bicycles are an example
of a moving object without battery, where tracking is more
challenging, as devices cannot feed off an existing power source.
Therefore many ways of tracking bike use have focused on devices
used by the rider (e.g. their mobile phone) or on attaching devices
with a long battery life to bikes [e.g. global positioning system
(GPS) trackers]. Whilst valuable in the absence of other solutions,
both strategies have limitations, since they require compromise in
terms of reliability and/or quality of data –the former because it
relies on people taking their phone on each trip made (with the
relevant application running), the latter because it relies on an extra
device being charged, switched on and attached to the bike ( plus,
choosing a setting that extends the device’s battery life results in
less data being recorded). E-bikes have an on-board battery that
can be connected to a monitoring system that tracks the bike
(rather than the rider via their phone) while not compromising data
quality for battery life (as is often the case in cycling research).
Several projects demonstrate methods for gathering sensor data
from cyclists. Dill and Gliebe [6] required cyclists to switch on a
IET Intelligent Transport Systems
Research Article
IET Intell. Transp. Syst., pp. 1–10
1
&The Institution of Engineering and Technology 2015
"This paper is a postprint of a paper submitted to and accepted for publication in
IET Intelligent Transport Systems and is subject to Institution of Engineering and
Technology Copyright. The copy of record is available at IET Digital Library"
GPS device for each trip (and mount it on the bike) and the data was
downloaded from the device after a period (of at least 7 days). Their
2007 study used GPS devices on normal bicycles, with 164
participants carrying a GPS device for (at least) 7 days each. The
research aimed to map where cyclists were riding their bikes in
Portland Oregon to assess the effect of different types of
infrastructure, such as bicycle lanes or paths, on bicycling [6]. The
team used the Garmin iQue, a GPS personal digital assistant that
was programmed to collect additional information from trial
participants about the trip and the related weather that had to be
put in manually for each trip. The device had to be attached to the
bike (unique mounting for each bike type) for each trip taken by
trial participants. The data was collected on the device’s memory
card. After collection of the device, the researchers analysed and
visualised the data and subsequently, trial participants accessed
maps of all their trips via an online interface where they were
asked to add more information for each trip. This required quite a
high level of involvement from trial participants (remembering to
take the device for each trip and attaching it to the bike, putting in
additional information about each trip, charging the device,
reviewing and annotating all trips via an online interface).
The Biketastic [7] and UbiActive [8] projects both used Android
applications that were installed to run on the participants’own
phones and used only the inbuilt sensors on the phone. The
Copenhagen Wheel [9, 10] took this further, by connecting the
riders’phones with Bluetooth sensors mounted on the bike.
Paefgen and Michahelles [11] used the Telex Picotrack GPS
monitor to track e-bikes, transmitting data via general packet radio
service. The BikeNet project [12] monitored a variety of sensors,
including video and pollution monitors, via a mobile phone, and
sent data via mobile network and WIFI. Paefgen’s Picotrack
monitor shows how a monitor can run independently using power
from the e-bike battery, however, these modules are limited to
GPS sensing only. The BikeNet project shows a good example of
using the mobile phone as the central part of a monitoring system.
Finally, the Campus Mobility project monitored e-bikes on a
university campus [2]. They used a small Android touchscreen
computer and GPS module mounted on the bike.
There are several public bike schemes that collect real-time data
about their usage but they rely on parking/charging stations.
E-bike hire schemes include those in Germany [13] and the
Netherlands [14], a forthcoming pilot integrating with a car
sharing company in the San Francisco Bay Area [15] and
institution-based system such as the one at the University of
Tennessee-Knoxville [16–18]. Many hire schemes log data when
bikes move in and out of parking stations, but most do not collect
data about the actual journey between stations. When analysing
the movement of bikes in public hire schemes, the trip data uses
the location and time of the station at the beginning and end of
each trip rather than a GPS data of the actual route taken, for
Fig. 1 SEMS was implemented on the two types of e-bike used in user trials, the Raleigh Dover (left) and the Velo-cité (right); both are used in low-step and
cross-bar versions
Fig. 2 SEMS was developed to collect real-time usage and sensor data, combining open source software and open hardware
IET Intell. Transp. Syst., pp. 1–10
2&The Institution of Engineering and Technology 2015
example, in an overview and analysis of data from 38 public bike
hire schemes across the world [19] and in more detailed studies of
specific schemes such as the public cycle hire scheme in Lyon
[20], Barcelona [21] and London [22]. Data analysis such as these
tend to assume the shortest possible route is cycled between both
stations, whereas other research shows that bike trips frequently do
not use the most direct route [6]. Therefore it was important that
SEMS records the entire e-bike journey in detail. Moreover, the
research fleet used by us needed to operate independent of
parking/charging stations because trial participants take ownership
of an e-bike for 6–8 weeks and use, park and charge them as they
please. Hire bike usage differs from personal bike usage, and our
trials were designed to simulate e-bike ownership and highlight
issues related to this, rather than those faced by a public hire scheme.
3 Design requirements
Scalability, replicability and modularity were overarching concerns
for the system design. These make it possible to grow our own
research fleet in the future or to replicate the SEMS for other
e-bike fleets, for example, in different locations in the UK, or in
other countries. This allows for the collection of comparative data
and lowers development costs for monitoring of e-bike fleets. To
allow for the scalability, replicability and modularity of the SEMS,
it was implemented with open source software and open hardware.
The additional design requirements for developing a system to
collect e-bike use data fall into three categories: experimental, data
and engineering. An overview of all design requirements in
relation to the projects and systems reviewed in Section 2 is given
in Fig. 3.
3.1 Experimental requirements
The key experimental requirement was that no user interaction was
needed (e.g. switching the monitoring system on and off, charging
it, transmitting data). This ensured that operating the monitoring
system is not a factor that could discourage trial participants from
using the e-bike or from logging its use. An autonomous
monitoring system also improves data quality as the participant is
not relied upon to charge batteries, or to remember to attach a
device to the bike when they depart for a journey. Additionally, it
minimises intervention from the researchers during the trial period
which results in a more natural context for the e-bike use.
The SEMS was designed to work with off-the-shelf e-bikes in the
low-to-medium price range so that the e-bikes used on trials reflect a
typical e-bike experience. The on-bike part of the system needed to
be weatherproof and robust to cope with vibrations and shocks from
road surface and bike handling. Finally, but most importantly, the
system needed to work safely and reliably.
3.2 Data requirements
The system needed to track each bike’s location (longitude, latitude,
altitude and time) so that each trip can be documented in detail (and
visualised on online maps). The altitude data is of particular interest
for analysing e-bike usage in relation to the terrain, to see if and how
usage (and use of the assistance) might change in relation to the
incline (steepness) of the route.
The SEMS needed to track how cyclists use the assistance on the
e-bike, including whether the assistance is switched on or off, and, if
switched on, reading which level of assistance has been selected
(low, medium or high). This allowed us to see how users use the
assistance in different ways, if specific types of usage emerge (e.g.
a user group that always uses the same setting for all journeys, or
another user group that switches the assistance on and off
frequently), if these patterns change over time for trial participants
(e.g. using less assistance over time), or how assistance use is
related to trip length, terrain and other variables.
The system needed to give real-time information so participants
can be given live online feedback. The real-time data also gives
vital feedback to the researchers for checking on the health and
status of the fleet. Additionally, this also allows us to make
selected ride information available to the wider public while (and
after) it is being collected by e-cyclists, for example, via social
media.
3.3 Engineering requirements
Power was a fundamental issue. To fulfil the requirement of
autonomy, SEMS needed to draw its power from the e-bike
battery and this meant that a key engineering requirement was
battery safety. The e-bikes in the fleet use lithium-ion batteries;
these can be permanently damaged if too much charge is drawn,
so a battery management system was needed to prevent the
batteries being drained below 3 V (volt) per cell. In addition,
SEMS should not draw so much power as to significantly affect
the range of the e-bike, so the monitor had to be designed to
conserve energy at every opportunity. The aim of autonomy also
meant minimising maintenance; the system needed to be protected
from the weather, in particular the corrosive sea air in Brighton,
the location of our study. Another concern was theft and
vandalism; to avoid undesired attention, the system needed to look
as inconspicuous as possible.
Finally, SEMS needed to be compatible with the two models of
bike in our e-bike fleet, the Raleigh Dover, and the Raleigh
Velo-cité, which we have in low-step and cross-bar versions, and
which are all from the 2011 collection (see Fig. 1). The Velo-cité
is designed to fit into the GBP1000 (Pound Sterling) price limit of
the UK’sCycle To Work Scheme that allows employees to get tax
and other benefits when purchasing a bike (to fit into this price
Fig. 3 Overview of SEMS design requirements (rows) in relation to the projects and systems reviewed in Section 2 (columns)
IET Intell. Transp. Syst., pp. 1–10
3
&The Institution of Engineering and Technology 2015
bracket, the battery needs to be bought as a bike accessory). This
e-bike model has a front wheel mounted motor, and a
rack-mounted battery. The Dover e-bike is a medium range model,
with a more advanced centrally mounted hub motor and battery.
The key differences affecting the design of SEMS were the
different drive systems and batteries. The Velo-cité uses a higher
voltage ten cell battery with a drive system made by ID-Bike, and
the Dover uses a lighter seven cell battery with a Panasonic drive
system. This meant that SEMS needed to run in two different
voltage ranges, and interface with two different designs of motor
control system.
4 Hardware design
The SEMS hardware consists of three main components: an Android
phone, an open hardware interface board and a custom power board
to connect the system to the e-bike battery. Fig. 4 shows the circuit
diagrams for the custom power board for the Dover bike and the
Velo-cité. All components are housed in a small water and dust
proof box behind the bike rack (under the saddle) as shown in
Figs. 5 and 6.
Fig. 7 shows an overview of the monitor system hardware. SEMS is
based around an Android phone, coupled with an IOIO board. The
IOIO is a low-cost interface board that connects to the phone via
universal serial bus. It allows connection with a large range of
sensors using analogue inputs and a selection of digital input/output
protocols. An Android software library allows communication
between board and phone. This combination gives the benefits of the
Android application programming interface (API) and phone sensors
(including GPS and accelerometer) along with allowing very flexible
hardware customisation with the IOIO. Crucially, the IOIO can be
powered by the bike battery, which can in turn charge the phone,
allowing the system to run continually without intervention.
4.1 Powering the system
The two bike batteries have different working voltage ranges:
the Dover from 21 to 30 V and the Velo-cité from 30 to 42 V. The
higher level represents the voltage level after a full charge. The
voltage must not drain below the minimum level or it will damage
the battery. SEMS connects to the e-bike battery in parallel, at a
place in the circuit before the e-bike’s drive system; this means it
can run when the bike is not switched on, but also means that it
needs to provide its own battery management system to prevent
over-draining. This management is provided by an under-voltage
lockout circuit, which cuts off power to the system below a preset
threshold, and restores it when the input voltage rises again. The
circuit is designed with a low-power LF442 op-amp to minimise
current drain. Both e-bike battery voltages are too high to power
the IOIO board which runs at a maximum of 15 V, so the voltage
is stepped down using an efficient LM2576 regulator. There are
two versions of SEMS, one for each model of bike. The two
versions are very similar, with small differences due to working
voltage ranges and assistance monitoring methods (see Section 4.4).
4.2 Phone selection
To choose a phone that would act as the hub of each monitoring
system, several Android phones were tested. Principal
requirements were low cost, boot-on-charge capability, GPS signal
quality, third generation (3G) connectivity and good battery life.
Boot-on-charge capability was fundamental to achieve the design
goal of autonomous operation for SEMS; a phone was required
that would boot up automatically (without human intervention) in
a situation where it had run out of battery (for any period of time).
Restoring the power supply (i.e. attaching a charged bike battery
to the e-bike) needed to automatically initiate the recharging of the
mobile phone, but also to reboot the phone and launch the SEMS
app. This was to cover situations where the participant may run
down the bike battery or remove it from the bike, causing the
phone to run out of battery. An initial design using mobile phones
that do not reboot created problems in real-world use, for example,
Fig. 4 Circuit board diagrams for the custom power boards for the Dover (left) and Velo-cité bike type (right)
Fig. 5 SEMS electronics are housed in a water-proof box that is mounted
under the saddle and fixed to the bike rack. No user interaction such as
charging or switching on is required
IET Intell. Transp. Syst., pp. 1–10
4&The Institution of Engineering and Technology 2015
when trial participants were not using the e-bike for longer periods of
time, for example, due to holidays or sick leave. Therefore, this
feature became the prime requirement for selecting a device. It
was challenging to find a low-cost Android phone with
boot-on-charge capability, and after some experimentation a phone
was found that, after rooting and alterations to low-level operating
system functionality, would do this reliably. This phone was the
Samsung Galaxy Ace 2 S6500. Further testing of this device
showed that the GPS quality, battery life and connectivity were
satisfactory for use in the project.
4.3 Deploying the hardware
SEMS was a rolling prototype, so rather than printing a printed
circuit board for the power board, the circuits were put together on
strip board. This would allow the addition of new sensors and
modifications over the project lifetime. The electronic components,
consisting of a phone, the IOIO board and the power board, were
placed in a environmentally sealed polycarbonate box, mounted
using custom brackets in the area behind the saddle. Durable
Teflon wiring was used to connect the components in the box.
Fig. 6 SEMS electronics, the IOIO board and the Android phone are housed in a water-proof box and are wired into the e-bike battery
Fig. 7 Hardware design of SEMS
IET Intell. Transp. Syst., pp. 1–10
5
&The Institution of Engineering and Technology 2015
Cabling from the bike is secured to the bike frame and enters the box
through a gland which is sealed with silicone. The box was grey, and
was chosen for its unobtrusive look. To draw power from the bike
battery, the positive and ground wires were cut immediately
following the battery connector, and rejoined with a three pole
connector, allowing SEMS to join the circuit at these points,
drawing power even if the bike was switched off.
4.4 Assistance monitoring
A key feature of the electrically assisted bike is that the rider can
choose to have assistance with pedalling. On both bikes, riders can
switch the pedalling assistance on and off, and are able to change
the assistance level that the motor will provide. This level of
assistance (low, medium or high) is chosen with a button interface
attached to the handlebars.
The two e-bike models feature very different designs for the
controller system, so different methods were used to monitor
assistance on each bike. The Velo-cité used the local interconnect
network bus serial protocol [23] to communicate between interface
and motor controller. The bus was connected to a UART port on
the IOIO, and software on the phone monitored commands from
the handlebar interface to detect assistance level changes.
Monitoring assistance on the Dover was less straightforward as we
had no prior information about how the interface communicated with
the controller. An experiment was conducted to explore the voltages
on the eight wires that run between interface and controller, and two
wires were found whose reading were indicative of assistance levels.
The voltages on these wires are monitored using analogue inputs on
the IOIO.
5 Software design
The software for the monitor system comprises two parts: the phone
software (Fig. 8) and the server software (a standard Linux system).
The client and server communicate over the internet using HTTP
protocol, connecting through a standard mobile data network. Data
from the e-bike monitors is received by PHP scripts running in an
Apache 2 server, and stored in a MySQL database. There is a web
interface for exploring the data, discussed in more detail in Section
6. The code is available from a public online repository [5].
The design requirements in Section 3 stipulate that the phone
software needed to run using as little power as possible, and send
sensor data back to the project server. To preserve the battery life,
the software runs as a background service, which keeps the phone
in a low-power sleep state for the majority of the time, with the
screen and communications services switched off. Every 25 s, the
phone wakes up and polls the accelerometer for 1.5 s. If motion is
detected, then the bike is likely being ridden, so GPS and
assistance monitoring services are switched on and the data is
logged. When motion is no longer detected, all monitoring
services are switched off and the programme returns to a sleep
state. To preserve power further, GPS data is stored locally, and
sent each 25 s in compressed form. The phone records longitude,
latitude, altitude and GPS accuracy, approximately once a second.
Assistance data is sent to the server every time there is a change.
Every 3 h, the phone checks in with the server to indicate that it is
still functioning, and sends information about the phone battery
level and whether the phone is currently powered from the bike
battery. This data is used to monitor the health of the bike fleet
and track any problems. A phone battery will last for around 4
days once the e-bike battery has been drained, with the software
continuing to send status messages to the server.
6 Online and social media reporting
The online system is used to analyse the sensor data and provide a
reporting framework. There are two facets of the system: one
private site for researchers where all data is accessible, and a
private site for participants to view their own data as they progress
through a trial (with secure log-in). Fig. 9 shows an example (fed
by test data) of the site that trial users would see after logging in.
The ride data can be accessed in real time, but also later on when
both riders and researchers can go through the archive of rides that
has been built up over time. Trial participants sign consent forms
detailing the collection and use of their ride data.
The system is built using Python and the Django web framework,
so as to give access to Numpy, a scientific computation package that
is used for analysis of the sensor data. The core of the system
analyses the GPS data, segments the data into separate trips and
calculates statistics about each trip: length, duration, start time and
end time. Trips are segmented using a time threshold, with gaps of
more than 3 min between GPS reports marking the end of a trip.
Trips are shown using a web interface, where the user can view
individual trips on a map created with the OpenStreetMap API.
Fig. 8 Design of the SEMS software
IET Intell. Transp. Syst., pp. 1–10
6&The Institution of Engineering and Technology 2015
A Twitter feed is used to share aggregated and anonymised
information about the trial bike use within the group of trial
participants and with the wider public (see Fig. 10). This is an
automated daily summary of how many trial participants have
used their bikes, how many trips they made and how many
kilometres they covered collectively.
7 System in use
In 2012 and 2013, four phases of 20 participants from two large
employers in Brighton trialled the bikes for a period of ∼2 months
each. Initial results of the 2012 commuter trials drawing on
surveys [24] and interviews [25] have been reported; future
analysis will include the GPS data collected by the SEMS.
Furthermore, trials with participants from the local community
took place in 2013, bringing the total number of trial participants
using SEMS to 92. The system had a maximum of 30 riders using
the system concurrently. Over the four phases, the system recorded
3645 trips, totalling 11,700 km, 775 h riding time and around
3,250,000 GPS data points.
The reliability of the system was tested through triangulation
between the different datasets that were collected for each trial
participant: surveys (before and after), interview or focus group,
odometer mileage and SEMS data. This reflects the mixed-method
approach of this interdisciplinary research project. As one check,
people’s assessment of the average mileage that they cycled on the
e-bike during a typical week of the trial (survey) was plotted
against the total mileage (based on GPS) recorded by the SEMS.
This shows a reasonable, though not perfect correspondence. As a
second check, the GPS readings from the bikes were assessed
against the interview data, to see if there was a correspondence
between the distance reading, and what people said about their
use. No obvious discrepancies were identified. While longitude
and latitude measurements were satisfactory, there were some
issues with variance in altitude data quality, as illustrated in the
example in Fig. 11. This could be corrected by using the
Ordnance Survey Terrain 50 dataset as a ground truth.
Three iterations of the SEMS were used on the trial during 2012
and 2013 (the SEMS described in this paper is the third and final
iteration) and this is reflected in the data collected. Iteration one of
SEMS was used by 29 participants and did not monitor assistance
(this was still under development) and used a non-auto-rebooting
phone (see Section 4.2) which resulted in some GPS data not
being recorded for some participants. The researchers received
automated alerts from the server if a participants monitoring
system was close to running out of battery (because the bike
battery had not been recharged for a time) and contacted the
relevant trial participants to ask them to either place a charged
battery on the bike or to keep a paper record of their trips, with
almost full compliancy. For the participants that used the first
SEMS iteration and where some of the GPS data was lost, a mix
of paper records (verified as described above) and GPS data is
used for analysis. Iteration two of the SEMS used an
auto-rebooting phone (see Section 4.2) and no significant GPS
data gaps occurred during the trial use by 20 participants,
Fig. 9 Trial participants can view their own ride data via an online interface with secure login
Fig. 10 Social media reporting of SEMS data via Twitter. Daily tweet
conveys the overall milage of the fleet to trial participants and the wider
public
IET Intell. Transp. Syst., pp. 1–10
7
&The Institution of Engineering and Technology 2015
evidenced through triangulation as described above. Iteration three
of the SEMS featured assistance monitoring and was used by 43
trial participants. The assistance monitoring was tested prior to the
trials on each SEMS through an experiment by the researchers
where they cycled a set course and used a set sequence of
assistance levels that were recorded on paper and through the
SEMS and subsequently compared. Fig. 12 shows an example of
assistance data collected with iteration three of SEMS.
The GPS data is analysed by using python code for cleaning,
summarising and segmenting into trips (see also Section 6).
Drawing on all three iterations of the SEMS, as a minimum, for all
participants, we have two headline figures: the number of days that
the bikes were used during the trial periods, and the total distance
travelled by the bikes. For sub-sets of trial participants (see
above), we have a much more detailed dataset. The assistance data
is recorded with a time stamp, and can therefore be analysed in
conjunction with the GPS data.
8 Discussion
The SEMS is a reliable way of collecting, analysing and displaying
e-bike usage data. Initial teething problems have been eliminated
through several design iterations, and the SEMS has proved stable
during several months of trials over 2 years. Some key issues that
have emerged during the design, implementation and trial usage of
the SEMS are discussed before contemplating future work.
The SEMS represents a trade-off between autonomy and data
quality on one hand and affecting the battery life of the e-bikes
on the other hand. The system, and especially the phone that is
part of it, runs down the e-bike battery over time, which affects
the battery life in different ways than usual e-bike use. A normal
e-bike battery runs out depending on how much it is used while
riding, with a number of variables affecting this, including how
much the ride assistance is switched on, what level of assistance
is selected, the length of the ride, the topography of the ride, the
weather conditions (especially the wind direction and speed) and
the weight carried on the bike (rider and cargo). The SEMS runs
down the e-bike battery in a different way, as it mainly
consumes battery over time, rather than only depending on the
ride. If a normal e-bike is left parked up for several days or
weeks, the battery level will be (almost) the same when the bike
is used again after the break in usage, as no bike battery is used
when the bike is stationary. An e-bike with the SEMS slowly
uses up the e-bike battery to establish whether the bike is
moving or not.
SEMS has two power consumption modes: when the bike is not
moving as little power as possible is used when checking for
movement with the accelerometer. During movement, more charge
is used because the GPS receiver is powered and data is
transmitted over the cellular 3G network. If a SEMS bike is left
Fig. 11 Example of GPS height data for journeys on a hill in Brighton
Fig. 12 Example of results from assistance monitoring, for journeys on the hill detailed in Fig. 11
IET Intell. Transp. Syst., pp. 1–10
8&The Institution of Engineering and Technology 2015
unused, with the bike battery plugged into the bike, the e-bike battery
will be slightly less charged if a rider comes back to the bike after
several days of non-usage, and after about 10 days of non-use, the
bike battery will be empty. After taking a break in using the e-bike
(e.g. due to holidays or illness), a trial participant has to recharge
the bike battery before taking the first ride. Alternatively, trial
participants can remove the bike battery from the e-bike, and plug
it back on the bike when resuming use (when the battery is not
connected to the bike, the SEMS cannot run it down). Trial
participants are made aware of this battery behaviour, and how it
differs from an off-the shelf e-bike. The vast majority of trial
participants are very happy to compromise battery life in return for
the ease of data collection. However, post-trial, many trial
participants also report that they would find full battery life very
useful if they were to own an e-bike themselves.
One of the unintended benefits of the SEMS became apparent
when one of the e-bikes was reported stolen by a trial participant
and the location of the bike could be determined by the researchers
via the online interface. After consultation with the police the e-bike
could be safely retrieved. The use of locational data for bicycle
safety is important for private owners and for fleet bikes.
9 Future work
There are several ideas for future extensions and iterations of the
SEMS. These draw on the experience gained from running the
system for 2 years, on feedback from trial participants and from
the companies involved with the commuter trials as well as from
discussing early findings with other researchers. One of the areas
we are particularly interested to develop further is the use of
SEMS and the e-bike fleet as part of a sensing network to collect
local environmental data. For example, we could attach sensors to
monitor noise pollution [26] or air pollution [27]. Another area of
interest for future development is to extend SEMS to measure
health variables such as heart rate monitoring and torque sensor
data [28]. This would represent a mobile health use case of SEMS.
An example could be smart e-bikes being used by those currently
physically inactive, as part of a health programme at work or
through a doctor (similar to current interventions with gym
memberships). The combined bike use data and health data could
then also support the economic case for this kind of health
intervention.
The system could also be developed to read detailed data about the
e-bike battery usage. This could be of interest to a range of
stakeholders, including e-bike manufacturers, battery
manufacturers and those purchasing or using fleets. Fleet
management of public or private e-bike fleets could also be part of
future developments. This use case concerns a fleet of public bikes
that are available for hire through public bike stations. Using a
public fleet with SEMS would enable crowd sourcing of bike and
sensor data at scale as it would engage with a large number of
bike users and with usage over time. Another potential use case is
the area of cargo e-bikes for urban goods delivery. In this case, the
SEMS could be developed to interface with the goods
management system and the battery use to calculate optimum
routes, to visualise goods progress in real time to customers, and
to calculate carbon savings compared with other modes of
transport. Another potential use case for SEMS is intermodal
transport. This would involve the integration with other modes of
transport via systems such as smartcards or mobile phone ticketing
and payment systems. Due to the modular design and the open
source nature of SEMS, some or all of these areas could be
integrated and several of the use cases could be explored in more
detail. This would work towards a toolkit of e-bike data that can
be combined as needed.
10 Conclusion
The SEMS is a stable platform for collecting, analysing and sharing
data about a fleet of e-bikes. The design, development and
implementation of the system contributes to understanding
e-bikes as a distinct mode of transport, and to conceptualising a
fleet of e-bikes as a distributed network, or an Internet of
Things. The current status of the SEMS fulfils the design aims
set out in advance, with the system running stable over extended
periods of time and with large numbers of e-bikes in use
simultaneously. The system works autonomously, resulting in
high-quality and real-time data about each bike’slocationand
the level of assistance chosen by trial participants. It is always
on and requires no interference from researchers that might
influence normal participant behaviour. The open source
approach in designing the SEMS makes the system replicable by
other research projects, following the details on our online
repository [5]. Future developments of the SEMS could include
research into health, environmental factors such as sound and
pollution, battery usage and fleet management, by implementing
the relevant sensors.
11 Acknowledgments
The Smart E-Bikes project is funded by the Research Council UK’s
Energy and Digital Economy programmes through the EPSRC grant
EP/J004855/1 and led by Dr Frauke Behrendt at the University of
Brighton, with Dr Sally Cairns (TRL, UCL) and David Raffo as
co-investigators, plus Chris Kiefer as Research Fellow. The
authors thank Lewis Boughtwood (for maintaining the SEMS and
implementing the SEMS Twitter feed), Dr Simon Walters (Senior
Lecturer in Electronic Engineering, University of Brighton),
Raleigh UK Ltd. They also thank Simon Ball (Senior Research
Scientist, Transport Research Lab) for the analysis of the
assistance data.
12 References
1 Cherry, C.R., Weinert, J.X., Xinmiao, Y.: ‘Comparative environmental impacts
of electric bikes in China’,Transp. Res. D, Transp. Environ., 2009, 14,
pp. 281–290
2 Mcloughlin, I.V., Narendra, I.K., Koh, L.H., et al.: ‘Campus mobility for the
future: the electric bicycle’,J. Transp. Technol., 2012, 2, pp. 1–12
3 GoPedelec: ‘GoPedelec handbook’. Technical report, Go Pedelec Project
Consortium, 2012
4 Behrendt, F., Cairns, S., Raffo, D.: ‘The smart e-bikes research project’, 2013
5 Kiefer, C., Behrendt, F.: ‘Smart E-bikes monitor system (SEMS)’. Available at http
://dx.doi.org/10.17033/DATA.00000016, 2013
6 Dill, J., Gliebe, J.: ‘Understanding and measuring bicycling behavior: a focus on
travel time and route choice’. Technical report, Oregon Transportation Research
and Education Consortium, December 2008
7 Reddy, S., Shilton, K., Denisov, G., et al.: ‘Biketastic: sensing and mapping for
better biking’. Proc. 28th Int. Conf. on Human Factors in Computing Systems,
(CHI ‘10), ACM, New York, NY, USA, 2010, pp. 1817–1820
8 Fan, Y., Chen, Q., Douma, F., et al.: ‘Smartphone-based travel experience
sampling and behavior intervention among young adults’. Technical report,
Intelligent Transportation Systems Institute, Minneapolis, 2012
9 Outram, C., Ratti, C., Biderman, A.: ‘The Copenhagen Wheel: an innovative
electric bicycle system that harnesses the power of real-time information and
crowd sourcing’, Monaco
10 Savage, N.: ‘Cycling through data’,Commun. ACM, 2010, 53,p.16
11 Paefgen, J., Michahelles, F.: ‘Inferring usage characteristics of electric bicycles
from position information’.LocWeb‘10, ACM, New York, NY, USA, 2010,
pp. 5:1–5:4
12 Eisenman, S.B., Miluzzo, E., Lane, N.D., et al.: ‘The design and implementation of
the E-BIKE physiological monitoring prototype system for cyclists’,ACM Trans.
Sensor Netw., 2009, 6, pp. 1–39
13 BMVBS: ‘Innovative öffentliche Fahrradverleihsysteme’. Technical report,
Bundesministerium fur Verkehr, Bau und Stadtentwicklung (BMVBS), 2012
14 GoPedelec: ‘Electric rental bike now inaugurated in the Netherlands’, 2011
15 Garthwaite, J.: ‘A Bay area experiment in electric bike sharing’, 2012
16 Cherry, C., Worley, S., Jordan, D.: ‘Electric bike sharing-system requirements and
operational concepts’. Bicycle Transportation, 90th Annual Meeting,
Transportation Research Board, 2011
17 CycleUshare: ‘cycleUshare –North America’sfirst electric bike sharing system’,
2013
18 Langford, B.C., Cherry, C., Yoon, T., et al.: ‘North Americas first e-bike share: a
year of experience’. Transportation Research Board, 92th Annual Meeting,
Washington, DC, 2013
19 O’Brien, O., Cheshire, J., Batty, M.: ‘Mining bicycle sharing data for generating
insights into sustainable transport systems’,J. Transp. Geogr., 2014, 43, pp.
262–273
IET Intell. Transp. Syst., pp. 1–10
9
&The Institution of Engineering and Technology 2015
20 Borgnat, P., Abry, P., Flandrin, P.,et al.: ‘Shared bicycles in a city: a signal processing
and data analysis perspective’,Adv. Complex Syst.,2011,14, pp. 415–438
21 Froehlich, J., Neumann, J., Oliver, N.: ‘Sensing and predicting the pulse of the city
through shared bicycling the bicing dataset’. Int. Workshop on Urban, Community,
and Social Applications of Networked Sensing Systems’UrbanSense08, 2,
Raleigh, North Carolina, USA, 2012
22 Wood, J., Slingsby, A., Dykes, J.: ‘Visualizing the dynamics of London s
bicycle-hire scheme’,Cartographica, 2011, 46, pp. 239–251
23 Denuto, J.V., Ewbank, S., Kleja, F., et al.: ‘LIN bus and its potential for use in
distributed multiplex applications’. Technical report number 724, 2001
24 Cairns, S.: ‘Electrically-assisted bikes: a way of mainstreaming cycling to work?’,
in: Velo City, Vienna
25 Behrendt, F.: ‘Sharing cycle rides on smartphones and city streets: towards
understanding the intersection of mobile media and electrically-assisted
cycling’. ECREA (European Communications Conf.), Istanbul, 2012
26 Maisonneuve, N., Stevens, M., Ochab, B.: ‘Participatory noise pollution
monitoring using mobile phones’,Inf. Polity, 2010, 15, pp. 51–71
27 Al-Ali, A.R., Zualkernan, I., Aloul, F.: ‘A mobile GPRS-sensors array for air
pollution monitoring’,IEEE Sensors J., 2010, 10, pp. 1666–1671
28 P. Chapko, A., Werth, D., Loos, B.F.: ‘A personalized and context-aware mobile
assistance system for cardiovascular prevention and rehabilitation’.
Lebensqualität im Wandel von Demografie und Technik, 2013 (See https://www.
vde-verlag.de/proceedings-en/453484032.html)
IET Intell. Transp. Syst., pp. 1–10
10 &The Institution of Engineering and Technology 2015