Ness's Chief Innovation Officer, Kuruvilla Mathew, gives his expert take on how Swarm Intelligence can be employed to fix traffic problems and prevent "Carmageddon"
Hello Everyone. I am Kuruvilla Mathew and I go by Mat. I am the Chief Innovation Officer at Ness Software Engineering Services. The topic for my session today is - Using Swarm Intelligence to Prepare for the Next Carmageddon.
In today session, I will be giving an dive into applying Swarm Intelligence and seeing how it can be applied to use cases while building Smart Cities.
I had worked on a Loss Prevention solution in Retail using Swarm Intelligence but let us dive into this use case in more detail …
Tweets
1. A single #IoT sensor or actuator isn’t smart, but a complete mesh can become one by applying Swarm Intelligence techniques #IoTSlam
2. Improve city planning activities by analyzing and applying Swarm Intelligence on #IoT data derived from Smart City Sensors #IoTSlam
3. We need disruptive techniques to solve our traffic flow problems in our cities by applying #IoT data and analyzing behavior #IoTSlam
4. Complement your traditional transactional data with #IoT data by applying Particle Swarm Optimization to gain actionable insights #IoTSlam
5. Understand natural collective behavior and apply it to understand how to apply it to #IoT solutions to solve congestions in cities #IoTSlam
What I will go through in the next 45 minutes is to give you a background of how it all started – It is always important to hear the story behind almost all IoT led initiatives.
I will give you a short story line to frame the context of this presentation and the session abstract but more importantly …
I will cover how Swarm Intelligences and particularly Particle Swarm Optimization could be used to understand traffic flow problems in cities or other places where there is commuter and pedestrian traffic. I will show some data sets that were captured and what insights could be derived.
So, lets jump into it …
Here comes the story…
Just a few years before, a big closure of I-405, dubbed Carmageddon. This resulted in a traffic jam that reached immense proportions.
Just in the past few months, Southern California experienced a 55-hour closure of the 91 Freeway, resulting in a 6-mile stretch that intersected two key highways. The closure was called the Coronageddon as it ran through the heart of Corona .
But the traffic problems and congestions are not isolated to one city or region they are everywhere.
While Carmageddon is an extreme instances of massive traffic congestion, but more commonly, we all deal with daily traffic problems that are created by early morning traffic as people get to work, school traffic, the lunch rush hour, and the all-too-familiar and stressful evening traffic.
Traffic flow patterns are studied by cities, but most use a low tech approach. They assign people to count vehicles as they pass through intersections at peak hours. This data is collected over a period of time, and then cities make decisions on whether to expand a road or add a traffic light or stop sign.
High-tech techniques do exist that can be applied to better plan and reduce traffic congestion. For example, already existing technologies can detect smart phone Bluetooth radios (for short range) and WiFi radios (for longer ranges) from vehicles as they pass through points where sensor detectors record the car’s presence. By placing sensor detectors at key locations along roadways, one can detect the general path of the vehicle as it passes through these points.
Now think of the possibilities of understanding common traffic patterns of thousands of vehicles in a crowded city. Having much greater transparency into traffic flow and congestion points could help city planners identify opportunities to smooth traffic patterns and more accurately plan infrastructure to support their cities’ growing needs.
I know that there are privacy issues to address but that will be a different topic altogether
Lets take learnings from nature with …
Swarm Intelligence
A single ant or bee isn't smart, but their colonies are. The study of swarm intelligence is providing insights that can help humans manage complex systems
It is the collective behavior of a group that is helpful in understanding and predicting behaviors based on some conditions
Applying Swarm Intelligence to traffic flow, city traffic congestion a number of challenges could be addressed.
As I mentioned earlier, Cities could capture the beacons from the connected cars and commuter smart devices to better understand behavior .
Cities then can implement better technology methods, such as swarm intelligence, to form a more accurate and complete picture of traffic flows, so cities understand where the real problems are and not have a person count vehicles.
Further to this, Applying analysis of this collated data could help optimize traffic flow and continually monitor, so adjustments can be made more quickly to avoid the next Carmageddon.
Let me explain what Particle Swarm Optimization is.
Couple of common definitions you will see of , (paraphrase) PSO is that it is a global optimization algorithm for dealing with problems, in which a best solution can be represented as a point or surface in an n-dimensional space. Hypotheses are plotted in this space and seeded with an initial velocity, as well as a communication channel between the particles.
In this flow diagram you can see how the vehicle (a particle) moves on the road and reacts to blockers (road closures, construction, school traffic etc.) and how behaviors change via visual indication, line of sight by one commuter (or more) in the collective.
I have some test data that I will show you later in this session , but let me get to the scenario
Let me take you to the scenario of Coronageddon. In spite of the couple-day closure, work continues on freeways and adjacent roads and has been going on for a number of months. This aggravates commuters and contributes to a fair share of road rage.
Thanks to google maps and the traffic view, I was able to capture the massive congestions that lingered for days. But this is a great place where, using PSO and the Bees algorithm, it is possible to understand and predict the behavior of the commuters at different times of the day. The changes of traffic patterns during the weekdays and weekends provide insights that can help city planners plan for future street and freeway closures
So, how do we do that? Installation of scanners along the streets can capture Bluetooth and WiFi beacons of commuters’ smart devices, as well as the Bluetooth beacons from vehicles.
We can begin to better understand the traffic flow by tagging the beacon from the vehicle and/or the driver and passenger in the vehicle. The effect of road closures that include streets and ramps can be understood by analyzing the vehicle/commuter between 2 points on the street.
Let me take you to another scenario where another type of algorithm like the Bees algorithm is applied
This Bees Algorithm could be applied to understand pedestrian traffic, especially near a school.
Applying the Bees algorithm to school traffic is an effective method to understand the traffic flow that is a combination of foot traffic and vehicular traffic. Traffic as a result of the start of a school day and dismissal will be an interesting pattern to observe.
For example, at this school, flow around certain intersection points had delays, but one intersection point was free flowing without delays.
Now let us get into the method of capturing the Scanner data
As described earlier, Scanners can be installed on streets, typically on street light poles, to capture the different types of beacons that are coming from the vehicle and commuter smart phones respectively. Vehicles built today have the electronics that are built into it that send out beacons that can be captured.
However, the number of Scanners will vary, with the amount of traffic that flows, the distances between them, reflection of signs, billboards etc. as they need to be placed in a manner that increases the chance of detection. The hardware and protocols will change as the IoT matures
Additionally, Having multiple Scanners also helps determine the vector of vehicle/commuter movement.
Bet you have questions on, how does Beacon data translate to real meaningful and usable data?
Well, Beacon packets come loaded with data – Machine Identifier (MAC), the Received Signal Strength Indicator (RSSI), Manufacturer and the Class of Device. So, it is loaded with rich information that can be used to get to meaningful insights.
A couple of things to determine is what gets captured at the scanners and computing the data at the Scanner or at the integrator depends on multiple factors. Power being one of them and the other is Compute. The RSSI is in decibels, so will have to be converted into a meaningful measure such as distance and more importantly the direction vector.
The direction vector is a bit tricky as it may require data from multiple scanners.
Let me show some of the data that is captured from a Scanner that captured the Bluetooth Beacon data and Wifi Beacon data
The captured data looks as indicated from the Bluetooth data logs. This reflects the Received Signal Strength Indicator (RSSI), including time stamp, vendor and a service tag identifier (ID).
A RSSI closer to 0 means that the vehicle is closer, and a higher value means the vehicle is farther away.
Using the class of device (cod) filter, you can isolate the captured frames that are most likely from vehicles.
In addition to Bluetooth data, another beacon captured is the WiFi beacons sent from Smart Phones and devices.
While the RSSI plays a key role in determining the distance for a given commuters smart phone (MAC),
it required some fuzzy logic to extract out the kind of smart device it is from the vendor data.
In the visualization what is interesting to observe is the behavior of the swarm.
One can see the changing behavior with time progressing. This can help determine a tipping point, whether it is start of rush hour or the end of one.
These insights are valuable in understanding commuter behavior with real data that can help city planners.
By using Swarm Intelligence algorithms, such as Particle Swarm Optimization (PSO), Bees Algorithm, city planners can create simulations to understand potential congestion challenges based on how vehicles and pedestrians navigate public spaces.
PSO is a good algorithm to apply to warehouses, as it helps them understand the behavior of each employee or a group of employees (beginning/ending of shifts) navigating out of facilities and getting on streets by walking, in vehicles, using public transport, etc.
Simulations using real data collected through this mechanism can help city planners determine potential traffic challenges at a highly-granular level—by street, intersection, freeway ramp, school area, etc. — to significantly improve the quality of empirical commuter data used in street flow planning and addressing existing congestion problems.
A quick brief and highlights of Ness Software Engineering Services
We have 10 Technology Innovation Centers across 6 countries
3,000 colleagues | Engineering team’s level of experience exceeds industry-average
Product Engineering rigor is at the foundation of our approach
Fully-integrated user experience design, platform development and data analytics services
I wanted to leave you with some interesting reading material in this area.
Thank you for your time and hope this session gave you interesting insights into how Swarm Intelligence could be applied to Smart Cities. I welcome questions offline as we do not have time. Hope you have a good remainder of the session at the Barbecode 2016 .
Thank you once again.