Last week, we gathered at the Paradigm Shift cafe in Koramangala, to learn about the location data infrastructure at Olacabs.com. The meetup was particularly interesting in the light of Ola’s recent move adding autorickshaws to their offering. Location is at the center of Ola’s business.
Vijayaraghavan Amirisetty, Director of Engineering at Olacabs, introduced how they collect data in real-time from cars fitted with smartphones. With over a lakh vehicles online at any given time, Ola’s primary challenge is to build an infrastructure to allocate taxis to customers quickly and reliably. Vijay highlighted some of the issues around collecting location data via GPS and cell networks. Even though both the technologies have matured since their inception, they are highly unreliable in various scenarios. Ola uses a combination of algorithms to build a reliable layer over GPS and network. One thing to note is that the smartphones are of variable quality and the system needs to work regardless of these metrics.
Even though Ola is using Google Play services as their location aggregator, in India, network is a bigger challenge. Quality varies from city to city and also reception within a city in unpredictable. Ola falls back to SMS, driver’s phone and a set of offline algorithms if the network is unavailable. Ola’s infrastructure is built using technologies like MongoDB, MySQL, Cassandra, Redis and Elastic Search. They are also exploring integrating web sockets and an experimental custom Android mod.
There was a lot of feedback from the audience specifically around why it is difficult for the drivers to locate the customer. Driver training is not an easy task – there are a lot of logistical and operational challenges. Vijay emphasised on the amount of work Ola does to improve the drivers’ experience with the whole process of on-boarding their cars.
Everything at Ola is realtime – why would anyone book an auto through Ola if they can just walk out and get one in less than a minute. They are continuing to improve and innovate to revolutionize transportation in Indian cities.
Autorickshaw photo CC 2.0 Spiros Vathis
To kick things off in 2015, we met at the offices of the Centre for Internet and Society (CIS), Bengaluru to map the unmapped/less-mapped settlements along the proposed Delhi-Mumbai Infrastructure Corridor (DMIC) project. The DMIC, a 1,483 km-long development corridor spanning several states in northern and western India, has been attracting a lot of curiosity and criticism from the national and international participants and observers. The project will have built a dedicated freight corridor, several industrial and logistics hubs, and smart cities at its completion. The project has been structured to be constructed in phases. The pilot project for an integrated smart city, Dholera Special Investment Region (SIR), is underway.
The quality of mapping in many regions relies on a very active mapping community, or a strong interest from a collectives and local networks. We think it is important regardless to map the assets that pre-exist around the proposed sites of developments. With this in mind, we decided to take a look at the areas earmarked for the Dholera SIR (Gujarat), Shendra (Maharashtra), Mhow (Madhya Pradesh), and Dadri/ Greater Noida (NCR). The evening began with Tejas introducing the DMIC project, the scale of new development, and the need to capture these changes for years to come on OpenStreetMap (OSM). Sajjad provided a rapid tutorial on signing up for OSM, and using the browser-based map editor. The party was attended by guests at CIS as well as remotely from Bangalore and Dharamsala.
As the party progressed, several guests ended up mapping roads, buildings, and water bodies in the Dholera region. Others chose to similarly map Shendra, and Dadri.
Last week at GeoBLR we discussed the issues around PIN codes. The most important questions were around the processes the postal system and also what are the issues around the availability of reliable spatial data.
Couple of weeks back, Nisha and I started putting together several questions that we would like to get insights on. We used that as the starting point for the discussions. The meat of the problem really is that nobody knows what the processes are and how to get that information.
Prior to GeoBLR, we met some people who are interested in the same issue and clarified a lot of things – for instance, we are now sure that some times a single post office can deal with more than one PIN code.
To get a sense how people felt about the PIN codes issues, we asked around. Some people don’t bother to use PIN codes for any substantial service other than sending post cards. As long as we are not able to tie PIN codes to geographic locations reliably, it’s not so useful. Everybody agrees that it has immense potential just because it’s the only part of the address that everybody gets right (most of the time).
We also started to brainstorm how to come up with a plan so that a group like ours along with several other partners could work together to attempt to crowdsource the issue. Read more about the plan and next steps here!
I’m excited to announce the first GeoBLR Sprint! The event is happening at The Center for Internet and Society on July 3, 6pm – 8pm. (RSVP)
During the July meetup, we are asking participants to bring their problems around maps and spatial data to the event. Some of Bangalore’s own data experts will be at the event, who will engage in a two hour problem solving exercise with the participants.
Have some map data that needs cleaning? Trouble with map projections or data formats? Looking for some data but not quite sure where to find it? Difficulty choosing colours for your map? May be we can help!
We encourage participants to get in touch with us prior to the event to talk about the issues that they would like to preset. Write to us on firstname.lastname@example.org, or post a comment on our Meetup group, or write to me (me at sajjad dot in). We will select couple of challenging problems and will recommend solutions for others.
See you at the event!
If you are curious to know more about GeoBLR and why we are doing it, I wrote about it here.