After product requirements get passed through user research and business development teams, engineering teams are responsible for assessing feasibility and practicality. They also determine the best approach in order to meet as many success criteria as possible. It is important to acknowledge that the technologies used—from the edge to the cloud—all have trade-offs that affect the cost, accuracy, and responsiveness of a system. In this post, we will take you through the primary considerations in architecting an indoor asset tracking solution.

Core Technologies

All indoor tracking solutions rely on one or several core technologies to provide accurate real-time location. Some examples include Bluetooth Low Energy (BLE), Ultrasonic, and RFID. Each of these has specific tradeoffs that dictate features such as signal range and penetration, which are important considerations when devising a solution for particular use cases.

In addition to the core tracking technologies, tracking solutions may be aided by supplementary data from sensors, such as accelerometers, gyroscopes, magnetometers and environmental sensors. The first three (accelerometers, gyroscopes, and magnetometers) can be used to estimate position through dead reckoning, which may be valuable in scenarios where precise tracking in complex environments is needed. On the other hand, environmental sensors, such as those that monitor temperature, air pressure, and humidity, can be used to determine details such as what floor someone is on, or what part of a space someone might be in.

Transmitting Data

Once you have determined which technologies to use and what data to collect, you need to decide on how you plan to transmit that data to the cloud. Depending on the volume of data being collected, there are two options:

1. When possible, edge computing is recommended to reduce both latency and cloud usage costs. There are also cases in which edge computing is necessary for a responsive system, such as when analyzing images or video.
2. Directly stream all data to the cloud for analysis.
   

This is typically achieved through gateway devices that communicate via LoRa or WiFi to the cloud. It is also possible to stream data from each node or sensor device directly to the cloud; however, this may overload the IT infrastructure or fail in crowded environments such as shopping malls.   

Transforming Data

Finally, you need to determine how to translate raw data into meaningful position information.

For this, there are three things that must be built:

1. Measurement Engine

Raw data tends to be noisy and requires some degree of processing before it is usable. What filtering and data transformation is needed? A good measurement engine is key to accurate position data.

2. Positioning Engine

How will you determine location? What level of accuracy is needed? How complex is the algorithm, and will it be affected by hardware limitations (e.g. smartphones or gateway devices)?

3. User Interface

What is being tracked, and where is it being tracked? How will location be visualized for the user? What level of interactivity is required?

“Gotchas”

Sometimes, despite all prior considerations, there are things that consistently emerge as pain points in system deployment. Here are some that industry experts see often and work hard to anticipate:

• Map inaccuracies
• Geo-referencing a large building properly is very hard. Oftentimes, there are image adjustments or distortions designed to represent wide areas from an aerial view, which affects where map features appear to be versus where they actually are.
• Signal attenuation
• Many signals are easily absorbed, blocked, or reflected by obstructions like human bodies or nearby furniture. These variations can cause inconsistencies or inaccuracies in the data.
• Change detection
• Because systems are often tuned up for specific environments and use cases, changes in room configuration or the accidental movement of beacons or other hardware can affect consistency of results.
• Physically crowded environments
• We recommend performing site surveys during off hours to reduce variability in data due to human bodies randomly moving around.
• Changing or incomplete requirements
• It’s cheaper to do it right the first time. This highlights the importance of good research to understand what the requirements and needs are of the use case.
  

The “full-stack” for indoor asset tracking spans across multiple technical fields—from hardware and firmware engineering, to cloud solutions architecture and data science, to front-end design and development. These considerations should be used as a guide and a tool when exploring approaches to building an indoor asset tracking solution.