A practical guide to RFID implementation in supply chain

Radio frequency Identification (RFID) as a technology has existed for a long time now. There are many applications of this technology in different domains and certainly a large portion applies to supply chain management as well. The applications of RFID in supply chain management has vast potential in improving effectiveness and efficiency. In this post we will cover a high level view of how RFID technology can be leveraged in the supply chain and specifically how we have leveraged this technology at Myntra to solve some complex problems. At Myntra we believe technology can democratise the way fashion is perceived and delivered. The below mentioned solutions are some in-depth work from the teams and most of it are currently deployed.

Before diving into the applications, let us understand the basics of RFID technology.

Fundamentals of RFID Technology

RFID has two major components, a tag (transponder — a microchip with an antenna ), a reader (a device with one or more antennas — to read data from the microchip using radio waves).

RFID Components

The RFID tag contains information about the object it is affixed. The tag can be of different sizes that can range from small as a miniature to larger sizes as well. There are two types of tags, active and passive RFID tags. Active tags are battery operated and can work in higher read range whereas passive tags are not battery operated, lesser read range but cheaper. Both active and passive RFID tags have various applications in the supply chain.

For this discussion we are considering only passive RFID tags

An RFID reader is a device that uses radio frequency to read information from an RFID tag. The data transmission works when an RFID reader (antenna inside the reader) sends RF signal to the surrounding area. The chip in the tag gets powered from the RF wave sent by the antenna, picks up the tag value stored . The tag reflects the signal while encoding the tag value along with it and goes to sleep. The reader, using an integrated or external antenna, acquires data from the tags, then passes the data to a computer for processing.

The reader passes the information to a computer and any business logic can be applied given a use case. There are many different types of RFID systems and it’s important to choose the right type of RFID system for a particular application.

RFID Advantages

Line of sight not required — RFID technology implies a greater Efficiency, Flexibility and Robustness because it does not need line of sight.
Bulk reading — Multiple and simultaneous reading of the tags are possible. This means we can read an entire bag of items in one go without needing to scan one by one.
Re-purposing tags —RFID tags contain a microchip with memory that can be rewritten ensuring greater accuracy and flexibility. This helps in re-purposing the tags in some use cases.
Durability — RFID tags that can work seamlessly in harsh environments.
Low Cost — Passive RFID tags are quite affordable and it can come as low as 2 Rs per tag or lesser. This makes it affordable in solving several supply chain use cases.

While every technology can be used to solve certain problems in this section, let us compare a few other technologies that comes to our mind closely when we talk about RFID. Let us compare their pros and cons and understand why we at Myntra chose RFID and not the below.

NFC is Near Field Communication. This technology is very close to RFID and in fact it is just low frequency RFID. While NFC also works passively the problem with NFC is you need to be really close to read the NFC tag. On the other hand the advantage of NFC is that it is low cost and instead of using a specialised reader one can use a mobile phone itself. However with the applications of supply chain the close proximity reading that NFC provides is not very useful.
UWB is Ultra wide band. UWB technology is very good when it comes to read accuracy and can support long range. UWB tags are generally active which means it needs battery power to operate. In effect, UWB readers / tags are more expensive. Given the nature of supply chain applications, at Myntra we prefer to use RFID, while we are still considering certain high value use cases such as tracking a trolley etc using UWB.
BLE is Bluetooth Low energy. This technology is widely used and BLE tags are again battery operated like UWB. Advantage of BLE is read accuracy and also read support available in mobile phones. However BLE is more expensive compared to RFID. One can consider BLE if the tags can be re-used. However, removing a tag from an existing package and applying it in another product, or reconciling these tags from products handed over to third party operations becomes complex and hence we chose RFID.

Applications of RFID in supply chain process

Now that we understand RFID as a technology let us look at where to apply this technology and what problems we can solve in the supply chain. Before we dive deep into this, let us also understand about supply chain process. The below figure captures the supply chain process/operations at a very high-level. This is a very simplified version of the supply chain process hiding certain levels of complexities. Let us try to understand a bit more on this process and how RFID can be applied at various levels. In the below diagram there are two flows

Forward — A workflow from order placement to delivery of the product to the customer doorstep
Reverse — A workflow from customer placing returns to product picked up and going back to its source location

Supply chain process

RFID Identity Creation

This is the process where products are brought into the warehouse. Typically these are called warehousing operations and managed by warehouse management systems. At this stage the items/products are moved from external sellers to a warehouse. The products are verified before moving into the warehouse. In the warehouse further processing happens such as stickering of each item, quality check of each item and moving to a storage area. These warehouses are temporary homes to these products and stay until an order is placed against them.

Process changes with RFID

A RFID tag is generated using a RFID printer for each product and affixed to the same
RFID tags are also mapped to GTIN/SKU ID (product identifier) of a given product in the backend

Note

In cases where the vendor already supports RFID tag, the receiving of products at this stage happens via RFID reader instead of opening the carton and scanning the barcode of each item. Using the power of RFID readability without line of sight and bulk reading capability, the operations team can simply scan the carton tag and then scan all the products present inside the carton without opening the carton. Post the validation via RFID the products are moved to the storage area

RFID printer and RFID passive tag

Hardware configuration

RFID printer with the capability of encoding and printing barcode information on the RFID tag

Bagging

Bagging is a crucial process in the supply chain where a set of products is put together physically in a bag (either cartons or crates or gunny bags (aka mother bag or master bag)). Bagging process is applied anywhere in the supply chain when you want to move goods from one point to the other. This means bagging happens in both forward and reverse flows while moving goods from a warehouse to logistics hubs to transport hubs to delivery centres and so on.

Now let us understand what happens at this stage and what are the problems that RFID can solve.

Bagging Steps

An operator keeps an empty gunny bag/crate/carton handy
Scans barcode of individual products that needs to be sent to a certain location
A system (typically an app running on a PC or mobile) validates if the product can be added to the bag and prompts appropriate error in case the validation fails
Operator keeps the product aside or adds the product to the bag depending on the validation result
Once the bag is full, operator prints a unique identity of the bag, sticks to the bag and closes the bag in the system
Post this the bag is opened only at the destination
A bag can reach its destination in one hop or in multiple hops

Problems in the above process

Misroutes — Sending wrong products to the seller/customer/location. Generally the system throws an error in such cases but if the operator ignored it, this can happen.
Missing products — Scanning the products to be added but not physically adding them. The system will show the product is present in the bag however the product is not.
Addition of products without scanning — Physically adding a product inside the bag without scanning the product. These are cases where barcode is damaged or operator simply did not scan the product

RFID Bagging

At Myntra we solved the above bagging problems using a RFID solution. We call it a 2-stage RFID tunnel. The solution is called 2-stage because we perform 2 stages of bag processing.

RFID validation of the product before adding to the bag. Biggest advantage with RFID is multiple products can be scanned in a single go and productivity increases by 3–5x at this stage
Continuous RFID validation of products that are added to the bag. Any user mistake will be caught using this process and immediately the bagging will be stopped

So how does it work

The 2-stage tunnel (as shown in the below diagram) is powered via one RFID reader and 4 antennas. 1 antenna is used to validate products before adding to the bag and three antennas are present inside the tunnel which continuously validates the bulk of products that are added into the bag.

Steps

Scan product on top
If valid piece, user can drop it into a bag inside unit
If invalid, separate manually
A bag scanner also keeps checking if top scanner has been by-passed

Engineering Design of 2 Stage RFID Tunnel

The figure below shows a cross section of a two stage tunnel that explains the dimensions of the tunnel as well as the antenna positions inside the walls of the tunnel and in outside extension. These antenna positions are tested and proved for the better efficiency of RFID tag detection on the items.

The figure also shows a typical metallic master bag frame that’s kept inside the tunnel which holds the masterbag with its mouth open. When the items are pushed from the top extension via the opening lid on top, it directly falls into the masterbag inside the tunnel.

Three antennas are placed inside the tunnel on the inner walls and one is kept on the bottom floor. 2 antennas which are on the left and right wall are mounted with the help of a hinge that can be rotated to get better detection. One antenna is kept inside the extension part on the top which is typically called the top scanner in our context.

One antenna on top constitutes one stage of scanning and 3 antennas inside the tunnel constitute the second stage of scanning, hence its a two stage tunnel.

Above figure shows the different motions that are configurable for two wall mounted antennas inside the tunnel. This different motion with locks help for better detection and efficiency in scanning the items inside the tunnel. The right antenna position is important so that RF waves emitted by the antenna directly fall on the masterbag. We experimented with antenna placements and came up with the most optimal configuration to ensure maximum readability.

Hardware configuration

RFID reader with 4 antennas
Adjustable antennas
Moveable tunnel

Results

Bag size — 35
99.99% scan accuracy

Handover / Receiving

In the supply chain goods are moved from one location to the other constantly. There is always a manual process while this handover/receive happens at every location. For example, when bags are moved from the logistics dispatch hub to the transport hub, it means the bag is handed over by the logistics dispatch hub and the transport hub is receiving the same. Post that the receiving hub processes the bag by opening it or further connects the bag to other locations.

Goes without saying, any mistake at this stage can cause huge leakage and misroute. It is also expensive to let go of errors at this stage because if the wrong bag is handed over at this stage it can reach a wrong location and the customer shipments will be delayed or lost eventually.

At Myntra we wanted RFID to solve the below problems in operations:

Reduce leakage / misroute
Validate whether shipments are intact inside the bag
Improve efficiency

One of the crucial aspects in the above problem is validating shipments inside a bag without opening the bag. This is where RFID helps. As we all know, RFID doesn’t need line of sight and has capability to read tags even if they are packed inside a bag. Further to this we created a conveyor based RFID solution that can do these validations on the go.

The RFID conveyor helped us validate bulk processing of items without opening a bulk entity (bag/trolley/bin/crate etc.). We also worked out a solution to segregate the bags based on validation. This solution brought huge efficiency improvements in the supply chain where thousands of bags need to be validated, moved across every hour.

We put together the below components to create a RFID conveyor

RFID reader
Metallic box
Conveyor with accept / reject arms
Conveyor Controller
Movement Sensor

Engineering design of RFID conveyor

The above figure captures the top and side view of the conveyor. The bags are loaded via the loading area which has a sensor. The sensor sends a trigger to the software application whenever a bag is loaded. The software on getting the trigger from the sensor puts the RFID reader in reading mode. After passing the sensor the bag reaches the tunnel which has the RFID reader fit inside it. The RFID reader is set to continuous reading mode and scans the entire bag along with its shipments as the bag moves through the tunnel. The data read from the bag is passed to the software which takes a decision whether to accept or reject the bag based on the bag validation rules. Once the bag moves out of the tunnel the software makes this decision and activates the reject arm via the conveyor controller in case the bag in invalid. The invalid bags move to the invalid bag area and the same is handled by an operator via standard operating procedure. All bags that are deemed valid moves to the valid bag zone beyond which the bag is moved to the appropriate docks without any manual intervention.

Conveyor Controller

As mentioned above, the conveyor controller helps in controlling the conveyor operation. This works as an interface between the software application and the hardware. There are three hardwares that a controller can communicate with:

Conveyor motor
Retro-reflective sensors
Rejection arm

The integrated conveyor controller can be used to communicate to the conveyor motor for the following reasons

Start conveyor motor
Stop conveyor motor
Adjust speed of the conveyor motor

The first two are action commands and need to be done realtime by the software and third is configuration command and needs to be done before starting the conveyor motor.

The conveyor controller can be used to communicate to conveyor retro-reflective sensors to sense object detection (of bags) so that appropriate decisions can be taken regarding the controller motor and buzzer/light according to business logic.

The conveyor controller can also be used to push the rejection arm and configure the speed at which the rejection arm is being pushed. When the application gives command to the conveyor controller, the controller should push the rejection arm to indeed push the master bag as a notion of rejecting it.

Conveyor controller connectivity

The conveyor controller can be connected to the application in the following three ways:

The PC containing the software is connected to the software controller directly via a serial port or usb port and the controller is connected to the conveyor system. In this way, the application calls APIs provided by the controller to control the conveyor motor, rejection arm motor and obtain information from retro-reflective sensors.

Conveyor process

In a session, multiple bags have to be checked for validity for a single container. So we can be sure that multiple bags need to be placed one after the other on the container to be scanned. For this reason, we need to make sure the following:

Two bags should not be considered as a single bag while reading, so appropriate distance should be maintained while placing the bags
Before reading the next bag, the processing of the previous bag in the backend needs to be over and an accept or reject decision should be ready for the previous bag.
Two or more bags should not be placed at the same time simultaneously on a running conveyor

To make sure of the first condition, the bags need to be placed leaving some space in between them. For example, if the distance between the tunnel and rejection arm is 5 meters and the conveyor runs in 0.5 m/s, then time for exit is 10 secs. This would mean each bag spaced 5 seconds apart gives a comfortable time gap between the bags.

To make sure the second condition is satisfied, the backend validation process has to be done as real time as possible and in our case we do this validation in under 200 msecs. As soon as the bag exits the tunnel the decision of whether to accept/reject the bag is made within a time t. If the processing is delayed more than t seconds then the software will reject the bag by default. The bag can be scanned once again by putting it in the beginning of the conveyor.

To make sure the third condition is satisfied, there is nothing that can be done at the start because there is no way of knowing how many bags have been placed on the conveyor as the sensor just does object detection. Therefore, if the presence of more than one bag is sensed inside the tunnel, which the software can detect based on the number of master bag tags read in one go, a buzzer is activated and all subsequent bags are rejected including the ones that were placed together on the conveyor. This would call for manual intervention and running the conveyor using the right standard operating procedure.

Hardware configuration

RFID reader with 4 far field antennas
Conveyor with accept / reject arm
PC connected to the tunnel to display results
Microcontroller connected to control the conveyor hardware
A sensor to indicate bags passing through the conveyor

Results

99% accuracy of bags in single pass
2–4 seconds per bag scan duration with 35 shipments in each bag

Tracking

Tracking is one of the most important needs in supply chain. Most of the tracking in supply chain happens via manual processes. If you see, even the operations like bagging, handover/receiving mentioned above still needs some level of manual intervention. However there are cases where there is a need to scan bags without any process involvement. Imagine you want to know how many bags are coming and leaving a transit hub without actually manually scanning any of the bags. One of the source of leakage and misplaced bags in lanes is in the transit hub. Transit hubs are locations in supply chain where bags are offloaded from trucks and temporarily stored before it hops its journey to the next location. Currently there is no visibility for bags that move via the Transit Hubs. It is certainly important to improve visibility for bags in the Transit Hubs and provide real time tracking and ageing alert based reporting to reduce leakages. In this section we are going to look at a solution to track bags without any manual intervention.

An RFID conveyor can do the job of tracking but there are space constrain in setting up a conveyor in some of the areas in the supply chain and the conveyor still needs some manual intervention.

We came up with an idea to implement a RFID gateway in the Transit Hubs and build a real time dashboard custom report generation to solve this. These gateway readers keep track of bags that are moving in and out of these transit hubs in various transport lanes. Let us look at how this works out.

RFID Gateway setup

RFID Gateway capabilities

An RFID gateway is nothing but a special RFID reader that is attached to any path way in an open environment
An RFID gateway continuously monitors for RFID tags that passes through it
An RFID gateway has the capability to understand the direction of movement of tags. This makes it easy to understand whether a tag is coming in or going out of a location.

It is very important to choose an RFID reader that has the capability to detect the direction of movement of tags for this use case because we need to know whether a bag is coming inside the cage or moving out. Not all RFID readers support this functionality. Please look at the hardware details captured below for further information.

RFID gateways are generally connected to the cloud via a compute unit attached to it so that the tags (bags attached with tags) can be reported to cloud
RFID gateway (or any other RFID reader) can raise an alarm with a buzzer connection via GPIO port (general purpose input/output).
Reports of tags read by the RFID gateway can be created using cloud backend

Structural Overview

The transit hubs generally have metallic cages where the bags are stored temporarily. These cages are specific to certain transport routes/lanes.

Our idea was to enable some sort of RFID scanning while the bags pass through these cages. We made the below changes.

Gateway Readers attached at the Cage Doors (top)
Buzzer Alarm/Light attached to the Cages
Each Gateway Reader is associated with a Cage ID
Cage IDs are further mapped to transport lanes

Behavior

When a bag is moved into a cage, the bag tag has to be read and status should be updated
When a bag is moved out of a cage, the bag tag has to be read and status should be updated
When a misrouted bag is entered into the lane, a buzzer/light alarm is activated and a notification is pushed to the dashboard

This solution helped us a lot in finding if a misrouted bag reaches in these cages. Once the misrouted bag is identified an alarm is raised to indicate the same to the operator. We built real time dashboards to display the bag and cage details so that the operator attends to this. This whole solution doesn’t need any manual intervention and can easily be applied in other areas of the supply chain as well, wherever bags need to be tracked.

Hardware Used

RFID reader with integrated antennas and capability to detect direction of tag movement
Thin client headless compute unit connected to the reader to get tag reads
Buzzer connected to the compute unit to raise alarm if wrong bags are passed

Results

95% accuracy of bags in single pass
Only bag scans are possible. Scanning shipments present inside the bag is not possible

Search

Searching for a product in a warehouse environment is no easy task. Given the complex processes in warehouse/logistics processing, the vast area and number of operators involved, there is always one or many cases where products/bags are left unattended or moved to wrong places inside a working environment. Such cases are very difficult to identify/correct manually. Imagine asking a person to go and search for a packed shipment or a master bag given an identifier in a heap of thousands of bags or in a huge storage area. The person has to search for each item using a barcode identifier to figure out whether it is present or not. This is almost impossible.

Now that we have RFID tagged every product and bag (as discussed in the above sections) the idea is to see how we can leverage RFID technology to search for products/bags. At Myntra, we evaluated a handheld RFID reader to achieve the same. Idea is for an operator to create a pendency of products that he/she wants to search for in an environment and walk around the area using a RFID handheld reader scanning/searching for the product/bags. Once the operator goes near the required target the mobile phone attached to the RFID handheld reader will start beeping and show product details indicating the fact that the product/bag is around. Further to this we adjust the reader mode such that the operator can pinpoint to the product and retrieve the same

Our Goal

To validate the hypothesis that a handheld RFID scanner can be used to reduce pendency related losses
To build a solution for searching tagged items/bags that are being stored at warehouse/delivery centres
Using a handheld RFID solution only, which means a user will have to go around looking for the items

RFID handheld use cases

We also had other use cases in mind like searching for products in trucks, using the same solution for hand over of bags to the sellers.

Handheld Architecture

Hardware

RFID handheld reader with BLE support

Software

Client / App

The RFID handheld reader supports command and control via BLE (Bluetooth low energy). This means a mobile device can be attached to this reader and can start receiving data from the read via a mobile app. We created an app based on react native and this module further integrates an android SDK provided by the manufacturer and via this SDK we use the reader functionalities.

The below picture depicts the high level architecture of the app.

RFID Handheld architecture

This implementation uses React Native, where the manufacturer SDK API is exposed to the react-native UI via the native bridge.
As all of the manufacturer SDK API is exposed already from the native bridge, any changes to the app only requires a JS bundle remote code push without the need to update the app.
Myntra has already built extensive libraries in react and react-native, we were able to add additional capabilities like authentication and authorisation without much effort by reusing them.

We evaluated the hypothesis of RFID tag search for products stored in two ways

Present in crates (as in supply chain mostly lose products/shipments are kept/moved across in crates)
Present in bags (Once the products are packed, the shipments are generally moved in closed master bags aka — gunny bags)

Results

We were able to easily locate products in both crates and bags in an dense environment

In case of crates, 5 crates (with 35–50 products in each) were kept in obstruction (1 crate on top and 4 on sides)
In case of bags, the bag was kept inside a pile of 50+ bags.

Observations

At a shorter distance (1 & 2 m), Crate and Bag readability are similar without any obstruction with more than 98% accuracy. But readability reduces slightly when other crates are kept as obstruction whereas it remains almost the same if the obstruction is another bag with items inside.

At 3 & 5 m, although the overall readability drops, it is better in the case of Bag with or without obstruction, both.

Note - Silk, liquid or metal items were not used for this testing.

The results of the RFID handheld search proves the capability of an extensive search solution that can be created using RFID technology. We were able to find the majority of the products/bags that are misplaced inside a location and retrieve the same. We have processes in place for operators to check for pendency items inside a location on a day-to-day basis and search, retrieve the same. This results in reducing major losses.

RFID Challenges

At this point from the above sections we should be convinced that RFID is indeed a useful technology in supply chain. However as with any technology, RFID also comes with certain limitations and operational challenges which makes the technology difficult to adopt. RFID can become a success story only when the below issues are mitigated in the right way. At Myntra we worked out various solutions and mitigation strategies to overcome this and a few are highlighted below.

All the above mentioned solutions are adopted in Myntra’s supply chain. Myntra is among the leading fashion, beauty and life-style e-commerce destinations and majority of the Myntra’s supply chain deals with fashion apparels, beauty & personal care, fashion accessories, home decors and the likes. The solutions we have crafted are based on the learnings from these product categories and other product categories may need some other solves.

Two fundamental problems with RFID

RFID does not work when attached to metals (Ex — if you attach a tag directly on a metal watch box then the tag cannot be read)
RFID does not work with liquids (eg beauty products like perfumes etc)

Mitigation

The above mentioned problems can be mitigated using some additional packaging material. For eg— we used paper wrap for these product categories and RFID tag was affixed on top of it. This enabled readability.

RFID Readability can vary depending on the environments.

For ex, tag readability inside a gunny bag/carton can vary significantly depending on the orientation of the tags and how the products are placed inside. The RFID tags may get rolled up and even touch each other in a packed environment like that.

Mitigation

Use a closed environment like RFID tunnel (metallic box) to read the bulk of products packed inside bags as the RF wave can get reflected inside the box
Use multiple antennas so that there is a higher possibility of reading

RF Leakage

One of the major issues we found in RFID adoption is the inability to control the RF signal. For ex, in a metallic box with RFID antennas we were able to read all products placed inside the tunnel however these boxes were not airtight and RF leakages were unavoidable. These leakages pose challenges as the RFID antennas are prone to read products that are placed even outside the tunnel as well. These are stray tags. One option is to reduce the power of the antenna (0.5 m) but this affects the readability of products that were placed inside the tunnel

Mitigation

Average out tag readings and filter tags based on the RSSI values of the tags (Receiver Signal Strength Indicator)
Use machine learning algorithm to determine the pattern of tag readings, their RSSI values and determine the final tags

Conclusion

RFID is a very useful technology, cheap and easily adoptable if one can understand the limitations and work around solutions accordingly. Specifically, in the supply chain the return of investment from RFID can be very fruitful.