Grab
Grab is a Singapore-based technology company that offers ride-hailing, ride-sharing and logistics services through its app in Singapore and neighbouring Southeast Asian countries.
They are a Gold Corporate Member of the OSM Foundation.
Grab - OSM Support and Usage
Grab would like to be part of the OSM community and help improve the map. Many places might not be accurate or identifiable on other forms of maps. Having these places improved by Grab will give richer information to OSM mapping contributors.
OSM Usage - Grab's data team contributes to OpenStreetMap's coverage of a number of countries. The team's mapping projects are documented publicly in the Grab Data Projects repository on GitHub. Grab's applications make use of OpenStreetMap data in few areas of South East Asia where commercial data providers have less extensive coverage.
OSM Data Mapping Guidelines
Data Modelling
The primary focus of work is adding missing roads and adjust alignment thereby enhancing the existing data in OSM.
In some cases, the existing road network has been mapped incorrectly, meaning that some features are missing or may be in-accurately mapped. We plan to use a variety of imagery sources to help map the geometry more accurately.
We always monitor the feedback from local mappers as local knowledge always precedes.
Improvements that will be addressed include the following:
- Adding missing roads
- Adjusting road alignment as per imagery and GPS traces
- Validate existing tags like- classification - Residential, Trunk, Motorway etc and attributes - Bridge, Tunnel, links etc.
- Cities might be revisited basis new improvements and or internal requirement.
Process Flow
Here is the process followed by Grab data team while modelling roads.
- Grab smoothened GPS traces(internal and non-downloadable file)
- Find the best imagery properly aligned per GPS traces and ground truth
- Use JOSM with GPS traces and best imagery as reference
- Visually identify missing roads and add them
- Correct classification and attribute errors in existing data
- Run validation and fix all errors and warnings
- Upload changeset with appropriate comments
Mapping Guidelines
The team's mapping projects are documented publicly in the Grab's Data Improvement Projects repository on GitHub.
Grab follows OSM's global and local policies, along with any other guidelines as appropriate.
Grab will also reach out to local OSM communities before beginning work, to gather feedback, suggestions and learn more about the data and tags used in that area.
Data sources and imagery we will use are listed here. Grab will research and use local tags and defer to local mappers, and also check the JOSM history on items to see why previous edits and changes were made.
Tools
The team will use JOSM for completing the task. JOSM has the OSM and Grab GPS traces (internal and non-downloadable file), and validation warnings for road networks and other possible issues that are important to prevent improper changes.
Changeset Comments
The team will provide changeset comments that are in compliance with OSM changeset guidelines.
Grab Specific Tags
With recent conversations with the Thailand community, the community and Grab data team is in consensus to repurpose import=* tag to import=grabremote to help the community differentiate between edits made by the Grab Data team.
Note: These edits are made using the traditional mapping process (there's no automation used), we are using an existing tag for easy maintenance and avoiding the creation of a new tag.
Quality Process
Below mentioned is the end to end process flow along with quality check gates at each phase
Phase 0 - Initial Validation
- Check for appropriate changeset comments and ensure they are in line with the scope covered in particular upload.
- Check if changeset boundary does not cover a larger area and/or multiple localities/cities/countries.
- Either case, as changeset cannot be reverted or modified, record the instances as feedback in the internal tracker, which will be used for further feedback to the respective production.
- Run validation queries, if errors found, reject the changeset without further evaluation.
Phase 1 - Quality Check 100%
Different validators other than production ops will perform a quality check. Follow the same process as that of production, while validating 100% of changesets marked as complete. Record the errors found and segregate basis Major and Minor errors -
Major Errors -
- Missing roads which are navigable
- Classification errors which are evident from satellite imagery on high priority roads - Motorway, Trunk, Primary
- Missing / Incorrect tags on above-mentioned roads
- Incorrect alignment (>3m) as per Strava heat maps (Do not micro align existing segments)
- Incorrect DoT(assigned / modified) on high priority roads
- Extra roads(created by GL users, we, however, do not delete existing data under any circumstances except if roads are duplicated) which are not evident from satellite imagery
- Incorrect connectivities which are not evident from satellite imagery
Minor Errors -
- Classification errors which are evident from satellite imagery on Low and Medium priority roads - Secondary, Tertiary, Unclassified, Residential and Service roads
- Missing / Incorrect tags on above-mentioned roads
- Incorrect alignment (>3m) as per Strava heat maps (Do not micro align existing segments)
- Incorrect DoT(assigned / modified) on low / medium priority roads
Meta Quality Check - 10% random sample
- A changeset will be marked accept when -
- Zero validation errors (With GL Id in history of the segment identified in validation check)
- Zero major errors
- No. of minor errors are less than or equal to 3
- Else Changeset is considered as reject.
Acceptance Criteria - Quality check 100%
- A changeset will be marked accept when -
- Zero validation errors (With GL Id in history of the segment identified in validation check)
- Zero major errors
- No. of minor errors are less than or equal to 5
- Else Changeset is considered as reject.
Feedback Mechanism
- Irrespective to changeset status, errors (major/minor) are documented in the internal tracker for further analysis
- Extensive root cause analysis(RCA) is conducted at an error category level to identify the root cause and derive appropriate corrective and preventive action (CAPA)
- Derived action items (AI’s) will be monitored for a period of 4 weeks to measure the efficacy.
- If the error rate is found decreasing, the implemented action item is documented as a best practice
- Else, root cause analysis and/or action items are found not effective, in which case analysis is conducted again with a different sample
- Error trend basis each operator is tracked to find out if there are repeated errors found.
- Appropriate actions are taken in this order -
- Performance action plan(PAP)
- Performance improvement plan(PIP)
- Refer to Human Resource Team (HR)
Error Detection
Grab will use various QA and data quality tools such as JOSM validation warnings and errors, as well as best practices on data quality from the OpenstreetMap Wiki page and LearnOSM.
Grab Data Team
Grab's data team contributes to OpenStreetMap's coverage of a number of Asian cities.[1] For any questions, please get in touch with Saikat Maiti or reach out to us on: geo.community@grab.com
Note: We are in the midst of updating this list, please look into Github for the updated list:
Below is a list of the team members supporting the mapping projects.
Chetan_Gowda (on osm)
Previous Team Members
Here is the list of previous data team members who currently are not a part of Grab.
General Feedback
For more information, suggestions, questions or general feedback contact our community lead Saikat Maiti Saikat Maiti (on osm) or email to geo.community@grab.com
Grab Local Map Operations team
To support the Grab data team, Grab has a local map operations team across eight regions in South East Asia. This team comprises individuals who help the Grab data team to get a better local context as well as support when there is a need for ground-truthing/verification. These teams also play a vital role in running multiple community activities and events across these regions. Here are the regions and their leads.
- Cambodia - Pimlada Veerapongawattana Pimladav (on osm)
- Indonesia - Ariek Wisnu Wibisono Jimbo314 (on osm)
- Malaysia - Nurmalina Zainal Abidin Malina Zainal (on osm)
- Myanmar - Pimlada Veerapongawattana Pimladav (on osm)
- Philippines - Golda Baoy GrabPHGolda (on osm)
- Singapore - Akbar Ali GrabSG_akbaralibakar (on osm)
- Thailand - Pimlada Veerapongawattana Pimladav (on osm)
- Vietnam - Nguyễn Dư Tuyết Nhi GrabVN_Nhi (on osm)
Note: We are in the midst of updating this list, please look into Github for the updated list:
Below is a list of the team members from the Map Operations teams.
Data Team Guidelines
In addition to the rules that apply to everyone in the community, here are the guidelines the Grab team reiterates and adds for themselves:
- We listen to the community. We are looking for your feedback on how to make a better map. For general feedback message Saikat Maiti Saikat Maiti (on osm) or email us to geo.community@grab.com
- Quality is paramount. We hold ourselves to the highest mapping standards as documented on the OpenStreetMap Wiki or as established as common practice in the community.
- Local knowledge first. Where in any doubt, the locally surveyed information prevails over remote updates.
- We disclose all ongoing mapping efforts on the Grab mapping repository.
- All full-time data team members will be listed here on this OpenStreetMap Wiki page and identified on their user profiles.
- Where possible we use public tools for coordinating work, allowing anyone in the community to participate.
Community Initiatives at Grab
Grab runs and supports some community events & programs in Southeast Asia, such as events like Mapathons, Map-a-Monday, Mapping Parties as well as participate in Humanitarian mapping efforts [2]
Geo*Stars is a programme for students to introduce them to OpenStreetMap, and Grab's Geo team, in Sept 2019, the programme was open for new applicants[3], but As of June 2022[update], that link is dead[4][5].
Thailand map errors report
After data quality issues were identified[6] in Thailand, an entry has been created to track and document errors by Grab's mapping team. There's a clean-up project in place that can be tracked in Github Issue 49.
See also
- Uber Another taxi-like company
References
- ↑ “Grab-Data/Grab Data Team on GitHub”. 17 Dec 2018. Retrieved 7 Feb 2019.
- ↑ “Grab and HOT Indonesia Hold Mapathon to Mitigate Impact of Natural Disaster”. Grab. 16 Dec 2017.
- ↑ “Join us as a Grab Geo*Star!”. Grab. 19 July 2019.
- ↑ https://www.grab.com/sg/blog/join-us-as-a-grab-geostar/
- ↑ https://www.grab.com/sg/search/?key=geostar
- ↑ https://techcrunch.com/2018/12/19/grab-maps-osm-thailand-southeast-asia/