Humans and geography

Driverless cars depend on Artificial Intelligence (AI) to function. Everything they do, from processing the data captured on their cameras and sensors, to interacting with human passengers, is underpinned by AI.

Here at Ordnance Survey (OS), we are researching how the cars of tomorrow will interact with passengers to receive instructions and keep people aware of what’s going on. One of the challenges we face is the way our language works. When we give commands, for example ‘take me to work through the park’, an autonomous vehicle must understand the different parts of the sentence i.e. ‘me’, ‘to work’ and ‘through the park’, the locations these phrases refer to – for instance, where ‘work’ is for you – and the spatial relationships – ‘through the park’ presumably means on a road crossing the park.

If a vehicle has been driving for an hour and needs to hand back control to a human, how and when in that process does it bring the person up to speed on where they are along the route? We’re exploring how the human-machine interface can support our navigational awareness in a safe and effective way.

In these ways we’re incorporating our knowledge of humans and geography with human interaction techniques to support the development of autonomous vehicles.

Nicknames for local places and landmarks are commonplace in the UK. Understanding exactly what location is being referred to can mean the difference between life and death. For example, in the event of an accident at sea or other emergency rescue operation.

Call handlers for emergency services don’t always have intimate knowledge of the landscape and local nicknames for the areas they serve. They may not know informal names of specific beaches or sections of woodland. That’s why we are building a directory of informal local place names.

Places and their boundaries are also often not clearly defined. For example, historical boroughs might no longer exist, but still be commonly used – the local recreation ground may have an informal name that locals knows it by. As such, our digital maps need to include different versions of boundaries. We need a deeper understanding of how different activities and narratives relate to different locations.

We are experimenting with crowdsourcing, updating vernacular geography data by allowing local residents to upload their own names for places, and are using data analytics to extract meaningful data from social media feeds. This can be useful for tweets that mention a major road accident or traffic congestion, for instance.

Virtual Reality (VR) and Augmented Reality (AR) offer new opportunities for geospatial data to be visualised in new ways. For example, when working in our streets being able to “see through” the road with an AR-equipped hard hat to see the buried infrastructure could help make sense of where to dig to minimise disruption.

We’re working to understand this balance between the information we need, and information overload – and the accuracy needs of the data and devices to assure that they are useful in practice.

In recent years, neuropsychologists have mapped out structures of neurons in the brain which store location information and the relationships between locations. We are interested in how the basic structures of the brain influence the way we learn about and navigate through our world. Can we design better navigation systems which are more in tune with how our brains understand the world?