Page 31 - FoodFocusThailand No.242 June 2026

SPECIAL FOCUS
3) The application of AI helps handle uncertainties the Fin-Ray Effect, which features a flexible finger structure that
in workpieces. The system can select grip points, adjust can automatically wrap around objects when contact pressure
gripping forces, and determine lifting directions based on occurs. In terms of adhesion, mechanisms include electro-
9
real-time data. When the operations of high-resolution adhesion and Gecko-inspired adhesion, which use flexible
sensors, camera images, and learning models are electrodes or micro-scale surface structures to generate attractive
integrated, grasping becomes gentler and more precise forces on flat objects without applying mechanical squeezing
and reduces waste caused by bruising. Furthermore, in force. Finally, granular jamming uses a silicone bag filled with
3
terms of robotic vision systems, artificial neural networks small particles; when air is vacuumed out, it stiffens and locks
such as CNN (Convolutional Neural Network), YOLO (V4, into a shape that conforms to the object. 3
V5, V7, V9), Mask R-CNN, and Inception v3 are used for
real-time image analysis. For example, YOLO is used in Post-Handling Quality Assessment
combination with 2D and 3D vision to locate coordinates After robotic handling, spectral imaging can be used to
of fruit stems, meat trimming points, and centers of mass assess product damage. This is especially useful for fruits and
to grasp items with balance and efficiency. 4 vegetables, where bruising may not be visible. Spectral imaging
In addition, the concept of Explainable AI (XAI) is can measure light absorption by water within fruit tissue (such as
beginning to play a role in enhancing transparency and at a 970 nm wavelength). This helps detect internal bruising in
traceability in chicken meat production processes. The products like strawberries and blackberries with higher accuracy
LIME (Local Interpretable Model-agnostic Explanations) than human visual inspection, and can help improve robotic
model has been integrated with the Inception V3 CNN to gripping systems over time. 10
classify the freshness of chicken meat. LIME functions Overall, AI-driven robotics is becoming a core infrastructure
by dividing chicken meat images into superpixels to for the modern food and beverage industry. It supports predictive
analyze which factors most heavily influence the AI’s quality control, gentle handling, waste reduction, improved safety,
decision-making, such as color, surface characteristics, or and enhanced traceability. Rather than being only an additional
other physical attributes of the meat. This helps humans layer of technological complexity, it is becoming a foundation
clearly understand and verify the reasons behind the for smarter, safer, and more sustainable food production
AI’s decisions, especially in cases where the system systems.
instructs the robot to sort and remove spoiled meat from
the production line. 5
Structures, Materials, and Mechanisms of More Information Service Info C002
Soft Robotics
Materials used in soft robotics are designed to imitate
the flexibility and movement of living organisms.
Common materials include silicone and PDMS
(polydimethylsiloxane), which are often used in Fluidic
Elastomer Actuators (FEAs) or Pneunets. These materials
are valued for their strength, flexibility, chemical stability,
and suitability for soft movement. Additionally, they
possess an elastic modulus close to biological tissues,
measuring approximately 0.1–10 MPa. 3
Furthermore, TPU (thermoplastic polyurethane) and
thin, soft PU that have received certification from the U.S.
Food and Drug Administration (FDA) are used. PETG
7
(polyethylene terephthalate glycol), which has received
food-contact certification , is also utilized. Moreover,
8
shape memory alloys (SMA) are applied to change shape
or adjust stiffness levels when stimulated by heat. These
materials may be used as single components or integrated
into hybrid/composite structures. An example is combining
shape-memory alloy wire with flexible silicone sheets to
create a gripping force that can envelop objects smoothly
and precisely.³
The gripping mechanisms of soft robots rely on the
principle of morphological computation, which reduces
the complexity of control software systems. It allows the
shape and material of the robot to adapt automatically
to the object. The core technologies used include Fluidic
Elastomer Actuators (FEAs) or Pneunets, which rely
on pumping fluid or air into internal chambers within
the silicone rubber structure. Crucially, an asymmetric
structural design causes the unit to bend toward the
object, which provides safety and allows for convenient
cleaning via Clean-in-Place (CIP) systems. 3
Additionally, there are tendon-driven mechanisms
that use cables or tendons pulled through 3D-printed
structures to control finger flexion. Another mechanism is
7
JUN 2026 FOOD FOCUS THAILAND 31

22/5/2569 BE 16:56
27-31_Special Focus_Ravis.indd 31 22/5/2569 BE 16:56
27-31_Special Focus_Ravis.indd 31