NASA Asks Crowd To Improve RFID Item-Level Inventory Location Accuracy On ISS
If you think crowdsourcing is just for small projects then think again. It turns out NASA has been active in coordinating collaborative innovation projects to tap into the dispersed knowledge of individuals around the world – aka ‘the crowd’ – to solve RFID inventory management challenges on the International Space Station (ISS).
Asset tracking on-board the ISS – items such as tools, scientific equipment, medical supplies, personal belongings, food and more – is mission-critical. Developing automated and highly-accurate systems is not only key for increasing efficiencies, but also for human survival. It is highly-complex and costly, if not impossible, to send up replacements for items missed or lost to long-term, deep-space missions.
Keeping track of things in space environs is also no easy feat. There’s limited space, rotating crews and, at zero gravity, assets need to be stored in containers or cargo transfer bags (CTBs) that look almost identical. When a container or bag is opened, multiple assets float away and risk being misplaced or lost. This is why NASA has been experimenting with RFID to improve the speed and accuracy of inventory audits and compliance checks – something we have written about previously. While traditional uses for RFID here on earth have been concentrated at box or palette level tracking through supply chains, the ISS project – known as RFID-Enabled Autonomous Logistics Management, or REALM – hopes to achieve location awareness of all objects at the item-level and eventually move to completely crew-free, fully-automated real-time inventory management.
To get a better sense of the magnitude of the equipment on board the ISS you can now see inside using Google Street View.
Improving The Item Location Accuracy Of REALM
REALM Phase 1 involved a UHF RFID system with 24 fixed antennas spread amongst 3 ISS modules (referred to as ‘instrumented modules’), approximately 3,200 tags on assets, and 100 or so marker tags on ISS internal structures. Many individual items were also contained within CTBs, all of which were also tagged. NASA’s internal results were able to locate assets with an average error of 1.5 metres, standard deviation of 0.5 metres, and maximum error of 3 metres within the instrumented modules. However NASA felt that if they could reduce the margin of error to 1 metre, it would lead the crew to a specific compartment or stowage area and virtually eliminate search time. They also wanted to determine how well the system could track items in non-instrumented (non-antenna) modules, since this would enable equipment (hence mass) reductions for deep-space stations and habitats.
So NASA decided to run a public International Space Station RFID Localization Challenge on crowdsourcing platform Topcoder to see if other algorithms could improve location accuracy and help astronauts narrow down where to look for missing assets. Contest entrants were given raw RFID data from the ISS and asked to detect the location of tagged items as accurately as possible.
The contest ran for five weeks, offered a combined prize of US$25,500 (ranging from US$8,500 for First Place to US$2,500 for Fifth) and received 25 submissions in the end. The winner was a US particle physicist who achieved a 28% improvement with a model based on weighted means of the antennae locations. In real-world conditions on the ISS this corresponds roughly to astronauts searching through half as much volume to locate a lost item compared to NASA’s algorithm. A comparable performance was also achieved by computer forensics expert from Brazil who used a complex combination of random forests (OK we admit we had to look this term up 🙂 ). Other prizes were awarded to entrants from China, the Netherlands, and Russia.
NASA was very pleased with the outcomes of the contest and novel approaches used. Not only did it result in solid improvements in location accuracy, NASA wants to further explore the random forest methodology and the approaches that delivered location results from the non-instrumented modules on the ISS.
Patrick Fink, principal investigator for NASA’s REALM investigations, noted. “There are so many different approaches one might pursue in an effort to improve the accuracy of the (RFID tag) location algorithms. However, these approaches require resources, and like most (government-funded) organisations, our resources are limited. So, we are very excited about the diverse set of approaches that can be rapidly evaluated for this challenging problem through crowdsourcing.”
If you are keen on learning more about REALM, NASA has also crowdsourced the production of a short video on Freelancer. The winner of the storyboarding contest was a psychologist from Australia (!) and the winner of the visual design was a 2D and 3D animation specialist from Germany.
Crowdsourcing Asset Intelligence
Relegen has long foreseen that crowdsourcing can result in significant improvements in asset data quality and the more efficient and effective management of assets. This is the process by which organisations can engage and leverage the dispersed knowledge of individuals enabled by the internet and real-time data capture provided by sensors and mobile devices. In a similar way, Relegen’s asset serialisation technology platform can be hosted in the cloud to enable multi-agent access. In doing so, it enables asset owners, operators, OEM suppliers and service agents to contribute to enterprise asset and risk data profiles every time they interact in the real-world – continually improving data currency, and quality, through-time simply as business occurs. Our solutions are also tag / sensor / infrastructure agnostic. It has been purpose-built to work with a wide-range of tagging technologies, including multiple technologies at once to deliver a secure, future-proofed, enterprise cloud and mobility platform. Not only can this capability improve item-level inventory management, it can also transform your business through the power connected assets, data and insight. If you’d like to learn more please reach out to us on +61 (0)2 9998 9000 or email@example.com.