Shoppers are already familiar with using the internet to help them navigate a store or direct them to special offers. Retailers also use the internet to keep the shelves stocked and deliver goods to customers. The next step is to use edge computing, image recognition and sensors for a ‘responsive retail’ experience.

“Smart retailers will map their IoT use cases to determine the right balance of connectivity cost, range, reliability, speed and features,” believes Pete Horwat, director of EMEA at Rigado. “As connectivity grows in complexity, and options continue to multiply, it is advisable to future‑proof an IoT solution by finding a gateway that offers a range of connectivity options, such as Bluetooth 5, Zigbee, Thread and Wi‑Fi,” he adds.

The company’s Vesta IoT gateway runs Linux and Zephyr OS, and supports Bluetooth 5, 802.15.4, Wi‑Fi and Ethernet. For secure edge computing, the company has added Canonial’s Ubuntu Core OS, which is designed for IoT and embedded devices running secure Linux app packages called snaps. Snaps have transaction update capabilities, allowing multiple applications to run on a single gateway. The gateways have a secure boot function and encrypted file systems to protect code and data integrity as well as preventing unauthorised code being loaded.

“A fully containerised edge OS, with security controls built‑in to both software and hardware guards against device security vulnerabilities,” says Horwat. “Furthermore, an integrated security‑as‑a‑service approach to updating and patching the edge, as vulnerabilities become known, helps keep retail IoT continuously protected in the field,” he says.

Last month, Rigado introduced a subscription service to manage and automate security. Cascade Edge‑as‑a‑Service includes the Cascade 500 IoT gateway, with Bluetooth 5, Zigbee, Thread, Wi‑Fi and LTE. It is an Edge Protect service providing automatic OS and security updates if vulnerabilities or exposures are discovered, while Edge Direct tools allow alerts to be set and to manage secure IDs and updates. Finally there is the Edge Connect connectivity and computing platform, with OS and snaps to maintain applications at the edge. It also offers easier connections to IoT sensors and beacons using API calls, says the company.

“Edge computing is critical for a number of applications,” says Horwat, “from reducing latency for local device control to pre-processing data, building and maintaining a true edge computing environment can easily consume 20% to 50% of IoT development costs.” He warns that this can lead to time‑to‑market delays, add costs and may lead to code being used that was not intended for production, creating a security and maintenance risk.

A pre-built IoT infrastructure used as a service allows developers to focus on building apps that add value. Retailers that apply this methodology save four to six months of development time, Horwat estimates.

Automated retail

Last year Intel announced a $100m investment in robotics, AI and IoT automation. The systems have been adopted by 60 stores, using 2,000 sensors for smart shelves, loyalty schemes, smart digital signage and inventory control. One example is in China, where online retailer JD has opened a bricks‑and‑mortar store that employs no shop assistants.

The D-Mart store sells a small part of the company’s range of electronics, groceries and clothing, and uses Intel responsive technologies encompassing edge computing, digital signage and point of sale (PoS) equipment.

Facial recognition and motion sensing are used to allow access to the store. Once inside, movement tracking can help to identify prime sites for product placement, and software updates maintain inventory levels. Biometric recognition is used for secure payment at the self-service checkout.

Chinese B2C e‑commerce is also embracing other emerging technologies with drone delivery. It is reported to have the largest drone delivery system infrastructure and has been testing robotic and driverless delivery services using an autonomous truck.

Customised chocolate choices

Image recognition can be used to identify customers who are part of a loyalty scheme. This has been used by North American artisan confectioner, Lolli & Pops, to allow assistants to recognise loyalty members in real time as they enter. Accessing those customers’ preferences allows analytics software to make personalised recommendations, tailored to customers’ known preferences.

Intel’s Responsive Retail Sensor is described as a store’s ‘nervous system’, connecting operating areas, collecting and processing data and using data analytics to monitor and control inventory. It has been used by Dutch clothing company G-Star in more than 20 stores to date. Sensors collect and process data, which is integrated with RIOT data analytics and retail management software to provide the retailer with insights into what lines are selling and need to be replenished. This data can prevent sales being lost through items being out of stock and avoid over‑stocking that can result in the need to discount goods to keep them moving.

Shelf stackers

Demonstrating how cameras and image recognition software can help to keep shelves stocked, Intel has collaborated with researchers at Carnegie Mellon University to develop AndyVision, an automated shelf stacker. In some shops suppliers are charged a premium to have products placed at prime retail points, such as aisle‑ends, or middle shelves. Using a map of the store the robot ensures that shelves and areas are stocked with the appropriate goods.

AndyVision is an alternative to using wireless RFID tags on products to track inventory. It combines image processing algorithms that run on a low power Intel processor to capture images. Although easier to implement than RFID tagging, Priya Narasimhan, professor of Electrical and Computer Engineering at Carnegie Mellon University, admits that the robot may be intrusive, mingling with shoppers while it captures images of the shelves.

The department is also looking into fitting cameras to some shopping trolleys to collect images unobtrusively as they are moved around, as well as using existing surveillance cameras or mounting sensors in ceilings and shelves around stores. The image‑based system can update brands daily via the cloud to confirm that product placement is correct and can also be used to update the robot’s store‑maps when new products are introduced, or if a product has been allocated too much or too little space.

On the move

While it is important to ensure that stock is in the right place at the right time, many shoppers also want to be wooed with special offers. Mobile coupons that are tailored to an individual’s preferences and buying habits can be delivered to a phone for use in the store.

Rob Hussey, market development manager at Honeywell for sensing and IoT, says: “These coupons are used as a link between the digital world and the physical store.” To create the link, many retailers are installing 2D image readers in kiosks in stores to read mobile coupons. At the checkout, or where personal information is scanned from an ID card, “it is important that the data is encrypted before it is transmitted over the internet,” says Hussey.

Connection to the cloud to facilitate data interpretation risks exposing personally‑identifiable information, whereas it can be extracted from an ID card inside a 2D image reader, for a more secure transaction, explains Hussey, citing the CM series compact 2D imager module for kiosk data operations. It has a 1D and 2D barcode scanner for phone screens or paper coupons. It is designed to be easily integrated into a kiosk, says Hussey, and can be mounted vertically or horizontally using a standard micro‑USB cable connector.

For North America, an EasyDL option extracts information from a driving licence, without transmitting it over the internet.

The customer profiling that can be achieved, if used wisely, can benefit customers with relevant offers based on their preferences, buying history or inventory available. “Devices such as the CM series Compact 2D imager module can be used to provide customer service at the kiosk that is almost on a par with the most trained customer service employees,” hopes Hussey.