The Indian Handmade Rug & Carpet Industry at a Crossroad - Part 2 Production

In the second part of my three-piece article on the impact of future technology on the Indian handmade rug and carpet industry, I explore how advancements can transform the production process.

Unlike large-scale textile manufacturing, which has easily adopted intelligent automation, handmade rug production faces unique challenges. Short production runs inherent to the industry, smaller-lot dyeing, and the intricacies of handloom weaving complicate every stage—from inventory management to quality control.

Inventory Management

Handmade rugs are predominantly "made-to-order," with minimal "made-to-stock" inventory. Fashion trends and unpredictable demand further complicate raw material inventory planning, which is already challenging due to the diversity of yarns—different materials, counts, plies, twists, and shades. Conventional inventory management approaches like MOQ, EOQ, and auto-reordering often fall short.

While the adoption of QR codes and RFID tags has improved inventory tracking, the future lies in visual inventory management. Systems like the "sortly" app, combined with OCR tools such as Google Vision AI and Amazon Rekognition, can differentiate between similar-looking raw materials and enhance accuracy. This visual integration allows access to details on smartphones and can help designers repurpose odd-lot yarns, creating new designs, reducing waste, and optimising inventory use.

NLP-enabled chatbot with vernacular voice assistance could further leverage this visual inventory management to aid stock picking. Users could query, "Do we have stock of the blue yarn used for Customer 123's rug in January?" or hold up a swatch against the camera and ask, "How much polyester yarn for this rug do we have in stock, and where is it stored?"—all in their local language at kiosks strategically placed within the warehouse.

AI-powered data analysis can guide the purchase department with intelligent stocking recommendations for some core ‘always-in-fashion’ raw materials while predicting seasonal fluctuations based on historical usage and pricing.

Dyeing

While controlling variables in cabinet dyeing has evolved with pH level control and PLC systems managing time, pressure, and temperature, the consistency of the dyestuff recipe remains an art. A robotic arm equipped with AI and micro-dosing dispensers, capable of dispensing as little as 0.01 grams of dyestuff—unachievable by humans—could revolutionise this process. Such a system could use an in-line photo-spectrometer to analyse, predict, and adjust the dyed yarn output in real-time, considering ambient light and drying conditions. This would offer precision and repeatability that is currently unattainable. It sounds sci-fi, but is imminently doable.

Weaving

While design serves as the brain of the rug, weaving is the heart! The uniqueness of a handmade rug lies in the blemishes and imperfections of the weaving pattern, reflecting the weaver's signature. While machines can replicate patterns with accuracy and speed, they may struggle to capture the nuanced artistry. Technology is unlikely and should not attempt to alter that distinctive quality. Moreover, there could be cultural resistance—the handloom industry is deeply rooted in tradition, and there may be reluctance to adopt technologies that could change the artisanal nature of the craft. Attempts at Robotufting within the Industry have only had limited success so far.

There are two key areas within weaving where technology can play a supportive role:

a. Pre-Weaving Tasks

Technology can minimize non-remunerative pre-weaving tasks, allowing weavers to focus on actual weaving and maximising output (wages are directly linked to woven output) to increase their income. This calls for flexible automation that supports human intervention and adjustments.

Some areas where technology can help:

- AI-Generated Warping Instructions: AI algorithms could instantly convert designs into warping instructions, saving time and boosting efficiency.

- Warp Beam Preparation: Collaborative robotic (Cobot) systems guided by AI could enhance consistency while not fully replacing human skill. Robotic arms with advanced computer vision that aids yarn characteristic identification (Material, Colour, Count) could expedite the warp beaming process with precision and speed, handling tasks such as knotting. For Power-looms, while the need is for warping automation that can create warps of 5,000 meters with speed, the handmade industry requires Cobots that can precisely and simultaneously create 10 warps of 200 meters each.

- Robotic Arms for Specific Tasks: Robotic arms with “end effectors” like precision gripping, similar to those used in surgical suturing or fishnet knotting, could be reprogrammed for handloom tasks like drawing, denting, and knotting.

- Computer Vision Systems: IoT-enabled sensor-based smart heddles and reeds can monitor tension and alignment in real-time across various loom types—Punja, Pitloom, or Framelooms—and can assist robotic arms equipped with requisite end effectors for the drawing-in process and ensuring correct thread placement. On-loom high-resolution cameras can periodically generate data for predictive analytics.

- Predictive Analytics: Image processing algorithms could predict warp outcomes, detect defects, and identify imperfections in yarn or color inconsistencies, enabling timely adjustments and reducing waste. Data generated from on-loom cameras check and predict pattern accuracy, predict completion time—specially useful for remote looms or long-duration projects like hand-knotted rugs.

b. Training

Technology should be leveraged to attract talent to the industry by promising better livelihoods. As skilled weaver shortages become pressing, innovative training methods are essential to preserving traditional techniques:

- Motion Capture Technology: Similar to proven culinary applications such as "Moley Kitchen" and "Chefee," which have replicated the hand movements of a chef, systems can record master weavers' hand and pedal movements, providing valuable insights. This data can be stored in detailed digital archives, preserving the intricate skills that define traditional weaving and form training datasets.

- VR/AR Training Programs: These can create immersive, effective training programs that teach traditional weaving techniques by simulating real-life weaving environments, allowing trainees to practice in a controlled, interactive setting using the training sets created by the master weaver and preserve the art of weaving. There are use cases in the revival of intricate crafts like pottery through robotic interventions.

Quality Control

QC in rug and carpet production involves addressing challenges like defects, color inconsistencies, pattern errors, and avoiding naturally occurring acceptable deviations. Traditional methods often struggle with the complexities of varying pile lengths, twists, counts, and intricate cut and loop pile combinations.

In 2016 at Ambadi, Kannur, even with knowledge support from IIITDM, Kancheepuram, we struggled to develop a computer vision-based system for identifying defects in handloom fabric (plain fabric, not a rug). It is remarkable to see how far technology has come in just eight years—today companies like Shelton Vision have developed advanced computer vision and AI technologies tailored for the textile industry, exactly along the lines on which we had envisaged.

Today, advanced technologies like AI-driven systems and machine learning models, specifically advanced object detection frameworks like YOLO (You Only Look Once), offer innovative solutions. Recent advancements in YOLO have achieved high mean Average Precision (mAP) scores, with AC-YOLOv5 reaching up to 99.1% detection accuracy. It can also create electronic maps of defects, detailing their location, size, and classification type, for future reference.

Rugs/Carpets with varying pile heights pose unique inspection challenges. YOLO's robustness to different scales and its ability to detect objects in a single pass make it promising for these materials. By training on diverse datasets, the system can accurately detect and classify defects even in fabrics with challenging textures.

I am aware that industry skeptics will raise eyebrows on costs and ROI in an industry dominated by SMEs. I will deal with that in my 4th section on implementation. In the upcoming 3rd section, I will deal with the impact of future technology on customer experience

The first part of my article can be found here