Physical AI: How Indoor Maps & Navigation Act as Spatial Intelligence

Even amid the unprecedented hype surrounding the AI boom, few subtopics have generated such interest and traction as the concept of Physical AI. In this guide, we’ll explore Physical AI and how it relates to spatial intelligence, specifically indoor maps and navigation tools.

At Pointr, we’ve been on the cutting edge of mapping and wayfinding technology for over a decade. As well as multiple patents powering our solutions and a roster of Fortune 500 clients to our name, we’re also proud to have been at the vanguard of AI developments in the space. Our revolutionary MapScale® engine was trained on thousands of manual mapping projects undertaken by our expert team and has enabled us to significantly reduce the time that map creation and digitization tasks take. Furthermore, MapScale® continues to outperform the leading commercially available AI models in its ability to detect, classify, and identify map objects during the digitization process.

In short, we know a thing or two about AI and how it relates to the challenges of mapping and navigating indoor spaces.

What is Physical AI?

AI is already having an enormous impact on the digital world. Physical AI refers to AI that can move beyond the digital and into interacting with physical spaces and locations.

This interaction covers everything from how data from physical spaces is gathered, to interpreting that data and making decisions with it, to how those decisions then go full circle to impact the physical space that first generated the data.

The degree to which the AI interfaces with the physical space varies significantly from use case to use case. Some need to take into account the most basic elements of the physical world - such as the presence of gravity and other forces - while others may take this as a given, and instead be working with more advanced data, such as that collected by digitized systems or sensors.

In short, Physical AI is any application of AI that extends beyond the digital and either deals with, or acts upon, the real world.

How does Physical AI differ from other AI and large language models (LLMs)?

It’s a common misconception that Physical AI models and solutions are entirely separate from what most people would consider ‘classic’ or ‘usual’ AI tools and large language models (LLMs) such as ChatGPT or Claude.

Rather, the core difference between Physical AI and ‘normal’ AI is the application it’s used for. As we covered in the previous section, any model or solution that uses AI to interface with any sort of physical environment can be considered an example of Physical AI.

This means that, while there are several advanced AI systems and models that exist specifically to operate as Physical AI solutions (for example, tools designed to digitize and interpret recordings from CCTV cameras, or models that take data from a building’s room booking and lighting systems and use AI to automate when lighting switches on and off for increased energy efficiency), there is nothing stopping tools such as ChatGPT from being used for Physical AI purposes.

How does Physical AI relate to indoor mapping and navigation?

As we’ve already seen, there are two key ways in which Physical AI interacts with the real world: taking data sourced from physical environments, and making changes to physical environments. Some Physical AI applications may do only one of these; others do both, taking information from the real world, interpreting and working on it, before transforming it back into tangible physical changes to the real environment.

In each of these scenarios, indoor mapping and navigation tools can play a crucial role.

• Scenario A: Physical AI making use of navigation and mapping data - There are countless ways in which data can be taken from an indoor navigation or wayfinding system and used by Physical AI. One of the cleanest is taking the numerous data points that any one user journey provides - where they began, their chosen destination, the time it took them to complete their journey, any significant deviations they made from their route along the way, and many more - and transforming them into actionable insights for the building managers. For example, instead of the building’s team having to parse hundreds or potentially thousands of user journeys, Physical AI could take the data and tell them if there was a particular spot or area where a disproportionate number of wayfinding journeys began, suggesting perhaps that that area is particularly confusing for users and could be adapted to improve this.

• Scenario B: Physical AI altering how people interact with their physical environment via maps and navigation - Just as the way users interact with a maps and wayfinding tool provides a treasure trove of potential data that Physical AI could then use to make material changes to the real world, so too can Physical AI impact maps and navigation tools. For example, take an electronic health record (EHR) system such as Epic MyChart which deals with patient appointments and can have maps and wayfinding embedded within it. If a certain physician has to cancel a day’s appointments, Physical AI can take their full schedule and distribute it among their colleagues to minimize disruption, and even take into account things such as each patient’s individual accessibility needs to ensure that the redistribution of appointments meets each patient’s needs. Finally, with this AI-powered redistribution complete, it can be fed directly to patients via their EHR, with refreshed wayfinding instructions, highlighted routes, and even suggestions for where they should now park in order to ensure the smoothest journey possible to their rescheduled appointment.

• Scenario C: Maps and navigation data feeding Physical AI, which then feeds back into the navigation and maps - Finally, we have the scenario in which the data from maps and wayfinding tools can be taken into account by Physical AI, and then in turn used in order to change the physical environment. One such example could be analyzing the patterns shown by map users’ live locations over time. This data could then be used to model ideas around building efficiency, for example in an office. While sensors have long been able to detect the presence of people in rooms and dynamically switch lights and HVAC systems on and off, by using Physical AI to analyze the broader patterns in which areas of the building are utilized at certain times, businesses could enhance their efficiency savings by avoiding situations where lighting and HVAC remain active for an entire wing or section to cater to a scattering of people who could be accommodated in a single area. Navigation systems could dynamically update to suggest users go to a certain bank of desks or a room where others already are, rather than sending them to an otherwise empty area and needing to switch all the lights and HVAC on. Meanwhile, maps could display occupancy data via color-coded rooms, and encourage users not to use empty areas if they can avoid doing so.

Why does adding spatial intelligence (via maps and wayfinding) matter to Physical AI?

For years at Pointr, we’ve helped businesses unlock the potential of spatial intelligence by providing them with highly accurate and detailed maps and wayfinding systems that help users - be they staff, patients, visitors, guests, or attendees - better understand the physical space they’re occupying.

The same is true for digital systems. Everything from room booking tools to HVAC monitoring systems to occupancy data to security features like panic buttons can be vastly enhanced by being combined with true-to-life spatial data. Users can better visualize the location of what data they’re looking at, improving their ability to interpret that data and the efficiency with which they act upon it. The same is true for Physical AI:

• Map and location data enhances context - Physical AI is wonderful at taking data and interpreting it. However, it isn’t perfect. When it comes to buildings and indoor environments, few things can improve the quality of AI outputs and the intelligence of the decisions it’s able to arrive at as much as giving it the enhanced context of a map. Doing so can help systems better understand the interconnectivity between different locations, the relative importance of one room or space over another, and more.

• Spatial data can act as a single source of truth - AI models running thousands of test scenarios on a digital twin to see the impact of potential changes, employees, visitors, or staff moving through the space, robots and other autonomous devices making decisions - having all of these use cases (and more) share a single, accurate map as their source of truth improves efficiency and saves money.

• Adding a human touch - For many Physical AI applications, the end aim is to improve how humans use, move through, and interact with a physical space. Providing mapping and wayfinding data (both the map itself, and the data for how users are navigating the space) adds a crucial element that Physical AI may otherwise miss; namely, that no two buildings are alike, and the intersection between a building and those visiting it is extremely difficult to predict. To avoid Physical AI tools thinking too robotically or prescribing unnatural ways of using the real world, there’s simply no substitute for accurate, detailed map and navigation spatial intelligence.

Industry-specific Physical AI use cases

As you will have gathered by now, the potential use cases for Physical AI are numerous; many, in fact, we haven’t even thought of yet. However, there are already many ready-made use cases tied to indoor mapping and wayfinding across a number of major industries:

Healthcare

• Changes within EHR systems - such as additional appointments being booked onto a patient’s calendar while they’re already at the healthcare facility - can be interpreted and reflected in the route a map user is told to take, which entrance and parking lot they’re told to use, and more.

• Autonomous systems, such as delivery robots, can better interpret the space around them, better understanding restricted zones such as secure or sterile areas and knowing to avoid them on their journeys.

Workplaces

• AI can look at office usage via employee calendars, and determine the days with limited in-office presence. For these days, it can then block certain areas of the office from being booked, enhancing in-person collaboration by keeping the workforce closer together and improving energy efficiency by ensuring that certain areas of the building can remain unlit and with systems such as HVAC turned off.

• Meeting rooms can be dynamically booked based on the size of the meeting, rather than simply letting the organizer select any room; if the meeting expands in attendees, it can be moved to a bigger room, with attendees alerted to the change and given preprogrammed wayfinding routes (or reminded about previous meetings they’ve attended in the same space to help jog their memory), and vice versa if the number of attendees falls.

Events and conferences

• Occupancy for a particular talk or session is monitored via live user blue dot positioning, and from the number of users who have yet to arrive but have earmarked that session either via a calendar option or via their current wayfinding destination. When a talk hits capacity, the venue is alerted and knows to open an overflow seating area, or to notify users yet to arrive that the particular session is already full and offer them routes to alternative options.

• Modern large-scale events frequently occur across a campus of buildings, with outdoor areas in between. By incorporating live weather data, the wayfinding system can update in real time to give users alternative wayfinding routes or even alternative destinations in the same building should their original destination involve going outside.

Retail

• Customers attempting to find directions to products that are not in stock can be dynamically redirected to similar items.

• Store associates on restocking runs can be routed more effectively, either by plotting the most efficient route through the store, or by giving them the most in-demand items to restock first (based on the map and search directory usage).

Are you ready to see how indoor maps and wayfinding can combine with Physical AI for your business?