As we strive toward an autonomous and connected future, the underlying technologies that will power this future will be critical. It is accepted that next-generation ITS and subsequently, future data sources, will form a key infrastructural foundation for the transit of tomorrow, whether that becomes autonomous vehicles, connected vehicles, mobility-as-a-service or some fusion of the three. The path from here to the future is yet to be determined and the Bluetooth probe data of today’s ITS could help point the way forward.

Data generated by Bluetooth devices is effectively used as a data source to provide detailed information for use in determining performance metrics, for example travel time and origin-destination. In a similar fashion, Wi-Fi data is another widely used option for collecting the core data set representative of road traffic from point-to-point. The commercial Bluemac data collection system has been deployed globally since 2008, collecting more than 25 billion data points in ten years and seeing success with state and municipal DOTs and traffic consulting agencies alike.

While Bluetooth Classic continues to be useful for probe data generation, the broader population of available devices is subject to the movements of the market. Noticeably, this has been reflected positively in the popularity of in-car entertainment systems designed to quickly pair via Bluetooth with drivers’ personal devices and negatively in phone manufacturers and users turning off their devices Bluetooth discoverability during general use or MAC address randomization by phones and tablets.

Enter Bluetooth Low Energy

Another option has roared into the market that has not generally been studied within the transportation industry, and therefore is neither widely understood nor applied, is Bluetooth Low Energy (BLE).

BLE occupies the same band used by both Bluetooth (which we will refer to as Bluetooth Classic) and Wi-Fi — they communicate on a frequency of 2.4 gigahertz, which has been set aside by international agreement for the use of Industrial, Scientific, and Medical Devices (ISM). This frequency is used by BLE for a distinctly different set of use cases between smart devices and operates completely separate from Bluetooth Classic and Wi-Fi.

BLE is used by a relatively new class of personal devices intended to facilitate a quick delivery of data at close range and has been implemented as the base communications means in fitness bands, smart watches, smart tiles, and iBeacons. Conversely, newer smartphones and tablets now ship with dual mode Bluetooth implementations enabling them to operate on both Bluetooth Classic and BLE simultaneously, and in some cases are core to advanced connection scenarios enabled by one or more major consumer device OEMs (such as Apple with AirDrop and Continuity). Meaning that in the near future the number of individuals donning at least one BLE capable device will continue to grow.

(Icons in graphic are from worldartme.com and pixcove.com)

ABI research estimates that, “2.3B billion phones, tablets and PCs will ship with Bluetooth in 2018 alone, with an additional 780M connected devices composed of 140M sports and fitness trackers”. The vast majority of these devices will support BLE; leading to the installed base of BLE devices worldwide being in the billions.

Understanding Bluetooth Classic

To understand BLE, we must first review how Bluetooth Classic works. Bluetooth Classic transmits data via low-power radio waves. Bluetooth uses a technique called spread-spectrum frequency hopping.  In this technique, a device will use 79 individual, randomly chosen frequencies within a designated range, changing from one to another on a regular basis. In the case of Bluetooth Classic, the transmitters change frequencies 1,600 times every second, meaning that more devices can make full use of a limited part of the radio spectrum.

Service-level security and device-level security work together to protect Bluetooth devices from unauthorized data transmissions. Security methods include authorization and identification procedures that limit the use of Bluetooth services to the registered user and require that users make a conscious decision to accept a data transfer. As long as these measures are enabled on the user’s device, unauthorized access is unlikely. A user can also simply switch their Bluetooth mode to “non-discoverable” and avoid connecting with other non-paired Bluetooth devices entirely. This means that a typical traffic probe detects Bluetooth Classic enabled devices only when they are in “discoverable” mode, typically providing enough captures equaling ~ 10% sample size on US roadways. One method used in the ITS industry to increase this sample size has been to capture the MAC address of devices set to “non-discovery” mode, something we have decided against at Bluemac Analytics in favor of the utilization of Bluetooth Low Energy data.

Understanding Bluetooth Low Energy

To mitigate interference in the 2.4Ghz band used by Bluetooth Classic and Wi-Fi, BLE also uses frequency hopping. Advertising intervals can be set to as fast as once every 20 milliseconds, meaning BLE often pings much more frequently than most Bluetooth Classic devices. Advertising is done sequentially on one of three dedicated advertising channels, allowing the discovery of devices available in the vicinity. Upon a connection request, the same channels are used for initial connection parameter exchanges. Once a device is discovered and connection is initiated, regular data channels are used for communication, as shown in the image below.

(BLE channel arrangements: engineering.stackexchange.com)

The channel plan consists of 37 data communication channels in addition to the 3 advertising channels used for device discovery. Advertising channels are allocated in different parts of the spectrum to provide immunity against interference from Wi-Fi devices.

BLE communications are designed to be shorter range, about ¼ of the distance capable of Bluetooth Classic communications conservatively stated at 10m. While MAC address randomization, as observed practiced by iOS devices, is certainly possible for BLE on phones and tablets, the vast majority of BLE connected devices were not designed with MAC address randomization thus advertise a consistent MAC address identifier.

Applying BLE for Enhanced Traffic Probes

Bluetooth Classic continues to be useful for probe data generation, especially as in-car entertainment systems are designed to quickly and efficiently pair to phones and tablets for entertainment use. However, the broader population of available devices is subject to the movements of the market, specifically phone makers or users turning off discoverability of their devices while out and about, or MAC address randomization by phones and tablets which slightly impacts matches used for travel time calculations with a bigger impact for origin-destination analytics.

What makes BLE especially useful for traffic engineering is the specific characteristics of BLE. Instead of having a discoverable mode, BLE devices advertise their presence in a much more frequent and efficient manner so that other devices can quickly connect and grab data for use by the end user. This means unlike Bluetooth Classic devices which may be limiting their discoverability, by design BLE devices are constantly advertising their presence across phones, tablets, smart devices, and beacons for proximity marketing and data exchange services between devices (such as AirDrop). Keeping those devices advertising is in line with both the experienced desired by consumers and core designs being delivered by the technology industry.

BLE is a relatively inexpensive technology. Unlike in-car entertainment systems which are not as widespread across vehicles, anyone with a fairly recent smartphone or tablet is likely advertising via BLE. With these cost advantages, BLE doesn’t suffer from the socio-economic penetration obstacles that are more prevalent with Bluetooth Classic technologies. This advantage is especially useful when deploying in rural areas, where often sample sizes are much lower with Bluetooth Classic.

BLE Application: Putting the Technology to the Test

Given that BLE and Bluetooth Classic are different but co-exist, both can be applied simultaneously in a properly designed system. With a system such as the Bluemac solution with BLE support, BLE can be applied to augment traditional Bluetooth Classic probe deployments as well as open up new scenarios to generate quantification of movements using the same proven methods applied for Bluetooth Classic. Bluemac Analytics’ industry proven practices with regard to privacy protection and filtering are all brought to bear for BLE support.

Increasing Unique Captures and Matches

While arguably the technique of logging non-discoverable Bluetooth Classic devices is potentially competitive; with augmenting via a BLE approach, a fundamental difference is using BLE the devices being logged are intended to advertise their presence, versus the non-discoverable approach where the logged devices are in a mode where the intention is not to be monitored.

For block-and-tackle vehicle travel time and origin-destination analytics, BLE has proven incredibly useful to supplement Bluetooth Classic captures. Bluemac devices have proven to be very effective at picking up BLE signals that reside in vehicles. As an example, on behalf of a state DOT we initially deployed a set of Bluetooth Classic only Bluemacs and upgraded this set with the ability to log both Bluetooth Classic and BLE signals a month after initial deployment. This resulted in an average of 4x to 10x increase in unique matches along any given segment, delivering a significantly richer data set as well as enabling the system to be statistically valid in ultra-low volume environments where Bluetooth Classic could be challenged.

The charts below show the unique capture and match rate increases for the Bluemac units that were upgraded with BLE capabilities. On average a 4-10 times increase in captures and matches were observed across the deployment, some as high as 30 times. While these results are based on actual deployments and live data, it should be noted that other deployment results may vary depending on environmental conditions and device placement.



The Ping-Pong Effect and Ultra Short Segments

Another benefit for travel time and origin-destination is the utilization of BLE’s shorter collection range. A specific challenge for both Wi-Fi and Bluetooth Classic are limitations on how short a defined segment can be between two endpoints to still achieve unique device captures that don’t overlap, AKA the “ping-pong effect.” This phenomenon occurs when two probe ranges overlap or nearly overlap: a single Bluetooth device can be seen simultaneously or in rapid succession by both detectors, resulting in impossibly short travel times. To combat this, inherent limits on minimum travel times exist to ‘filter’ much of this behavior out of normal use. The consequence of this method is that short segments therefore see a drastic reduction in valid matched devices.

However, BLE’s shorter range has shown to be an excellent tool for countering the downside of closely placed probes. With one client, Bluemacs are providing effective data via two probes which are line-of-site functioning on a ¼ mile long segment of highway. While this can be achieved with Bluetooth Classic, additional work goes into design and on-site work is required compared to BLE which essentially works “out-of-the-box.”

The customer chose to install BLE capability on both units in this segment to gauge the effect on their data validity and number of matches at any given time. The below figure shows the median pre-BLE and overlaid post-BLE numbers on a single day in red and blue respectively. The delta in raw matches between the two probes nearly doubled over the day and the standard deviation of the median decreased by nearly 10%.

The Path Forward

Applying BLE for Multimodal: Track the Journey, Not the Mode

BLE has also shown to be effective for capturing pedestrian and cyclist movements. Since BLE devices tend to be personal and are typically worn or carried at an individual level, the potential exists to be able to track multimodal travel across transit mediums. Recently we conducted a study on Tilikum Crossing in Portland, OR, the first major pedestrian/cyclist/transit-only bridge in the United States. This proved to be the perfect proving ground to demonstrate the effectiveness of BLE for non-motor vehicle movements. By applying proper filtering and validating with traffic camera counts, matches were able to be differentiated and banded. Through the study measuring one hour of peak commute time, the system generated segment matches for over 40% of the bicyclists and pedestrians that completed the segment!

The same is true of the light-rail trains and city buses that use the bridge – these movements could be quantified using a BLE-enabled Bluemac system similar to the figure below.

These findings demonstrate a great application for BLE – true multimodal tracking. Pivoting from logging a vehicle; instead, anonymously log a person on their journey as they transfer between different modes of transit. With a properly set up Bluemac deployment with BLE, these modes of transit can be determined at each step in the journey and a richer origin-destination data set can be generated. Individual privacy is protected with the same secure methods Bluemac has always used.

Why are we spearheading this field? Because we are a research-driven company committed to applying and delivering new technologies and solutions to the transportation industry and the greater public they serve.

Bluemac Analytics believes this is just the beginning of BLE’s applications to ITS. We continue to invent and prototype to take BLE to the greater market as we move towards the autonomous future and the scaling of mobility-as-a-service. As always, we’re excited to share these endeavors with interested parties.

Contact us today to find out more.