Doug Rinckes, Google

Philipp Bunge, Google

Public discussion list: open-location-code@googlegroups.com

In much of the world, street addresses are poorly defined or non-existent. This causes many problems, such as being unable to receive deliveries, direct emergency services, or manage disaster relief. Solving this problem in the past has focused on naming streets and numbering houses, but for even a small area this can take years and cost significant amounts of money. An alternative solution that uses tablets and smartphones to encode location information into easily exchangeable codes would be immediate and free. This paper proposes a new standard to generate such codes that can be used to communicate precise location information from person to person, or between person and computer.

In large parts of the world, homes and buildings either do not have street addresses or have addresses that cannot be precisely located without specific local knowledge.

This is due to a variety of factors, including:

• Streets that are unnamed or have been renamed a few times

• Unofficial roadways or settlements (e.g., slums such as Kibera in Kenya and Dharavi in India)

• Newly constructed streets whose names are not widely known

• Areas that have the same or similarly named streets in close proximity

• Locally used names of streets that differ from the official names. These could be completely different names or abbreviations of the official names

• Unusual orderings of building numbers - non-consecutive or not aligned with the street.

Often, the most precise and easily accessible street maps are online. Updating online mapping services with the road layout is relatively easy, but adding street names takes detailed local knowledge. This can take years, and until it is complete the maps are of limited use.

It is estimated [farvacque] that half the world’s population live in urban areas and that half of those have no street address. In many cases, people are fully connected to and accessible by the internet but without an address they are disconnected physically. They have laptops and mobile phones. They can, for example, order goods online, but are unable to receive deliveries at home and must use a delivery address, or poste restante [poste_restante] service. Typically, if they know a package has been sent to them, they have to repeatedly travel to the nearest post office to check if it has arrived. Businesses have similar problems with making and receiving deliveries, and delivery services need to allow for extra time to locate destinations. This effectively limits many businesses to operate only in their immediate locality. Lack of addresses also causes problems with public health management and other public services.

There are cases where street addresses are not useful, even where they are available. For example, locations inside parks may be hundreds of meters away from the nearest address. Even for buildings that do have unique addresses, there is no practical way, for example, to specify the location of different entrances. This particularly affects large entities like factories, hospitals and sporting facilities. Shopping malls also typically have a single address, but may have several entrances in addition to each shop having its own location.

Many European countries have small villages where the streets have no official names and the buildings are not numbered. There may also be buildings far from the nearest road, such as huts and refuges in mountains. In addition, street furniture (e.g., benches, vending machines, fountains) don’t have street addresses but it would still be useful to be able to refer to their location.

Three approaches are usually used to provide a location in these circumstances. The most common solution is to provide simplified directions instead of an address. This results in addresses such as:

• 11th km of Old Road from Heraklion to Re [arolithos]

• in front of old civil engineering lab, Oke Ado Road, Ogbomosho [ogbomosho]

• Up the winding road, 600 m from the bridge, Past Lazy Dog, past English Bakery on the left at Hotel Mountain Dew, Manali, Himachal Pradesh [manali]

These directions rely on detailed local knowledge and are difficult or impossible to geocode algorithmically. If the directions cannot be followed, there is no alternative other than to ask around and hope to find a guide.

Where there are nearby street addresses, an option is to use the best available street address. This can lead to a building having an address of the nearest named street, which could be some distance away.

Finally, latitude and longitude provide an exact location, are used internally by GPS and satellite navigation devices, and are sometimes printed on paper maps. However they are rarely seen on city or street maps and are difficult for people to use. They consist of long and complicated numbers, have different ranges (-90 to 90 vs. -180 to 180) and need to be used in a specific order. (To express a reasonably precise position requires between 14 and 18 characters.)

A system of encoding location information into a short and an easy to use code would solve these problems. As smartphones become cheaper and more widely used, they could provide a way to convert easily communicated identifiers to locations. Such codes could be used and exchanged over email, phone, in print and handwritten.

We believe that these codes need to meet the following requirements for widespread adoption.

• Easy to use

• Complete

• Flexible precision

• Indicate proximity

• Cultural independence

• Functions offline

• Easy to implement

• Free

Easy to use: The codes must be short enough to be remembered and used. This means that they need to be shorter than latitude and longitude and about the same length as a postal code or telephone number. The symbols used to make up the code should not include characters that can be easily confused (e.g., 1 and I, 8 and B, 2 and Z, 5 and S etc.)

Complete: The codes must have enough information on their own. Extra information (such as street number, locality or country name) could be helpful, but should not be required.

Flexible precision: The location precision required for a sports field is less than that required to locate a utility meter. Locations should be expressed to an appropriate degree of precision. For example, the location of an apartment building may only need a precision of +/- 10 meters. But locating a smaller house may require a precision of only three or four meters. The codes need to support a range of precisions for different situations, and the precision of a code should be visually apparent.

Indicate proximity: It should be possible to determine if two codes are near each other by looking at them. Ideally, it should also be possible to determine direction, and even to have a rough estimate of distance.

Culturally independent: The symbols that make up the codes need to be widely available and recognisable. Codes should avoid including profanity or other words in multiple languages.

Function offline: The codes will be used in both built-up and rural areas, so must be able to be created and decoded without a data network. This also applies to users who are roaming or who live in areas where data networks are expensive.

Easy to implement: Software libraries should not be challenging to implement. Creation and decoding of codes should not depend on a single provider. Codes should be discoverable and usable by anyone with the appropriate hardware and software. Codes should be available without having to apply to a standardization body or a central provider.

Free: Adoption of the code should not require a license fee or be otherwise impeded by licensing or patent restrictions. Additionally, the codes should not depend on a single provider for their continued use.

Open Location Code is a new way to express location that meets these requirements. It is shorter than latitude and longitude because it uses a higher number base. It uses a number base of 20 because:

• In base 20, 10 characters can represent a 14x14 meter area suitable for many buildings

• Using a number base of 20 makes some calculations easier

• We could identify a 20 character subset from 0-9A-Z that doesn’t spell words.

The characters that are used in Open Location Codes were chosen by computing all possible 20 character combinations from 0-9A-Z and scoring them on how well they spell 10,000 words from over 30 languages. This was to avoid, as far as possible, Open Location Codes being generated that included recognisable words. The selected 20 character set is made up of "23456789CFGHJMPQRVWX".

Note on terminology: The characters from 0-9A-Z that make up the significant part of an Open Location Code are referred to as "digits". Additional symbols used for formatting are referred to as "characters".

Open Location Codes are encodings of WGS84 latitude and longitude coordinates in degrees. Decoding a code returns an area, not a point. The area of a code depends on the length (longer codes are more precise with smaller areas). A two-digit code has height and width[height_width] of 20 degrees, and with each pair of digits added to the code, both height and width are divided by 20.

The initial pair of codes identify a cell from a 18 x 9 grid covering the Earth, where each cell in the grid is 20 degrees by 20 degrees. The first digit of the code identifies the row (latitude), and the second digit the column (longitude). Subsequent steps divide that area into a 20 x 20 grid, and use one digit to identify the row and another to identify the column. Defining codes in this way allows for the proximity and direction from one code to another to be determined visually, and for codes to be truncated, resulting in a larger area.

The large rectangle in Comparing areas of four and six digit Open Location Codes is the Open Location Code 8FVC (1 degree height and width). The smaller rectangle is the code 8FVC22 (1/20 degree height and width).

A 10 digit code represents a 1/8000° by 1/8000° area. (At the equator, this is approximately 13.9 meters x 13.9 meters.)

A 10 digit code will be precise enough for many locations. However, in areas where building density is high (such as informal settlements, semi-detached houses or apartment blocks), such an area could extend over several dwellings. A 12 digit code would be less than 1 square meter. An 11 digit code would be preferable because it is shorter, and a slightly lower precision area could be acceptable.

From 11 digits on, a different algorithm is used. The areas are slightly larger but the advantage is that the codes are shorter.

The area of the 10 digit code is divided into a 4x5 grid, where each grid cell is identified by a single digit. The digit for the cell containing the desired location is added to the code.

Using a single grid refinement step, we have an 11 digit code that represents a 1/32000° by 1/40000° area (roughly 3.4 by 2.7 meters at the equator).

The first approach (where a pair of digits is added for each step) provides codes that can be visually compared, or alphabetically ordered to determine if they are close to each other. The second approach allows the code area to be refined using only a single digit. If the entire code was generated using the second approach, it would result in codes that could not be reliably compared visually.

10 and 11 digit codes provide the necessary resolution to represent building locations. Other lengths are also valid.

We are accustomed to providing different levels of detail in a street address depending on who we give it to. People far away usually require the full address including the country. Within the country, we may give state-level information, while people in the same neighbourhood may not even require the city name. Information can be omitted because it is obvious from the context.

Similarly, by providing a locality name as context (that can be geocoded to a latitude and longitude) the leading digits of an Open Location Code can be omitted. The original code can be recovered using the provided context, or even with different but nearby coordinates.

For example, the Nairobi Youth Sports Organization and Information Centre in Kibera, Nairobi, has the Open Location Code "6GCRMQPX9G". Using the location of Nairobi, the code can be shortened to "MQPX9G".

This method of shortening the code only requires that the location of the code, and the location of the place, are within approximately 40-50km of each other. It works because the correct location is the nearest one that includes the short code, "MQPX9G".

This means that the Nairobi Youth Sports Organization and Information Centre can use the full, global address "MQPX9G Nairobi, Kenya". Within Nairobi, a person can just use "MQPX9G" and find the correct location.

When combined with the town or neighbourhood name, most people will only have to remember from four to seven digits of their code.

A "+" symbol is inserted into the code after the eighth digit. This performs two key functions:

• It allows us to recognise shortened code fragments such as MQPX+9G. Because we know that the "+" is after the eighth digit, we know that there are four digits to be recovered for this code.

• It allows us to distinguish four or six digit codes from postal codes.

But this means that we have a problem if we want to represent the 1x1 degree area 6GCR. The solution here is to use zero, as a padding symbol, giving us 6GCR0000+. Zeros in Open Location Codes must not be followed by any other digits.

Open Location Code has some imperfections, driven by usability compromises or the encoding methodology. The key ones are listed here.

• To prevent the codes including words, some letters are not used. For example, A and B are not used in the codes. The codes W9 and WC are next to each other, but this isn’t immediately obvious

• The character set is defined in Latin characters. We have considered defining different character sets for different languages, but there can be problems identifying the language if visually similar characters are used. For example, it is difficult to distinguish the latin "H" from the cyrillic "Н". Although latin characters may not be the first choice in many areas, it is probably the most common second choice throughout the world

• Code areas distort at high latitudes due to longitude convergence. The practical impact of these disadvantages are not significant due to the low populations at the north or south poles, and the ability to use codes representing small areas to approximate point locations

• Code discontinuities at the poles and longitude 180. Codes on either side of the 180th meridian, although they are close, will differ significantly. Similarly, locations at the poles, although physically close, can also have significantly different encodings. The fact that there are no significant population centers affected means that this is an imperfection we are willing to accept

• Open Location Codes cannot exactly represent coordinates at latitude 90. The codes for latitude 90 would normally have an area whose lower latitude is at 90 degrees and an upper latitude of 90 + the height of the code area, but this would result in meaningless coordinates. Instead, when encoding latitude 90, a code with an upper bound of 90 degrees is produced. Normally, the upper bounds are not included in the area. This means that we cannot exactly represent latitude 90 in a code. We are willing to accept this shortcoming since there is no permanent settlement at the North Pole.

Refer to Open Location Code Specification.

• [poste_restante]Post restante (French: lit. post remaining or general delivery) is a service where a delivery is made to a post office that holds the package until the recipient calls for it.

• [height_width] "Height" and "width" are used as a shorthand for north/south latitude distance, and the west/east longitude distance.

• [farvacque] Farvacque-Vitkovic C, Godin L, Leroux H, Verdet F, Chavez R 2005. Street Addressing and the Management of Cities, World Bank, 2005

• [arolithos] Google Maps listing, Arolithos Traditional Cretan Village, Arolithos 71500, Greece, https://goo.gl/maps/CXI37

• [ogbomosho] Google Maps listing, Lautech Old E.E.E Laboratory, Ogbomosho, Nigeria, https://goo.gl/maps/Dnnin

• [manali] Google Maps listing, Open Hand Cafe Manali, Manali, Himachal Pradesh 175131, India, https://goo.gl/maps/bh5bo