Delta Robot Selection: Parallel-Link Pick-and-Place RFP and FAT/SAT for Thai Plants

Operators picking products off a conveyor and arranging them into trays or cases by hand: this is still a common sight in Japanese-affiliated food, pharmaceutical, household goods and electronic component plants in Thailand. When you try to automate this process, delta robot selection quickly becomes the question, because the usual candidate is the ceiling-mounted, high-speed parallel link robot (delta robot). In practice, however, many teams end up almost settling on a model based on the large “XX picks/min” or “cycle time XX s” figures printed in the catalog.

This article makes one argument. Delta robot selection is not decided by the catalog’s maximum speed or maximum payload. Catalog figures are measured on test motions such as “25-305-25”, and the conditions differ from manufacturer to manufacturer. The capacity you actually get on a real line is determined by 4 factors: (1) the pick time with your actual workpieces and your actual hand, (2) conveyor tracking and vision processing, (3) the required orientation rotation (on some models the payload changes considerably depending on whether a rotation axis is used), and (4) washdown and hygiene conditions. That is why the RFP (request for proposal) should require “a demonstration with our own products, hand, layout and target capacity”, and the FAT (factory acceptance test) and SAT (site acceptance test) should confirm that “when the robots run continuously at the target capacity multiplied by a margin factor, missed picks, drops and misplacements stay within the allowable range”.

To help production engineering managers and plant managers assemble the material they need for internal decisions, this article works through the following 5 steps in order.

1. Define the target process and required capacity (product flow rate [pieces/min], variation, peaks, placement patterns, and how many robots share the work)
2. Decide the workpiece and hand (vacuum or gripping, pick and release time, whether rotation is needed, workpiece weight + hand weight)
3. Narrow down the models (number of axes, payload including whether a rotation axis is used, work envelope, hygienic specification and protection rating, mounting method)
4. Design vision, conveyor tracking and load balancing across multiple robots
5. Safety, hygiene and acceptance (guarding, ISO 10218:2025, ISO 13857, cleaning procedures, FAT/SAT)

Note that the flow rates, robot counts, headcounts, amounts and payback periods in the capacity design example and the model calculation in this article are all assumed figures set up for this article. They are not industry averages, survey results or manufacturer prices.

Why Delta Robot Selection Is Coming Up Now in Thai Plants

There are 3 drivers behind plants in Thailand considering automation of the tray-loading process: labor costs, the export environment, and the global trend in robot adoption.

First, labor costs. According to an announcement by Thailand’s Ministry of Labour (July 17, 2025), the minimum wage in Bangkok became 400 baht per day for all industries from July 1, 2025. Minimum wages differ by region, and the same announcement also lists provinces at 372 baht and the lowest provinces at 337 baht. Because the effective date may also differ by notification depending on the province, you need to check the amount and effective date for your own location individually. Regarding a revision in 2026, there are media reports that as of October 2026 there is no sign of an increase, but this is not a primary source. In any case, the cost of staffing the tray-loading process should be viewed not as the minimum wage itself but including social security, bonuses, overtime, and the effort of hiring and training.

Next, the export environment. According to media reports of an announcement by Thailand’s National Food Institute (NFI), Thailand’s food exports in 2025 were 1.51 trillion baht, down 8.1% year on year. January-February 2026 was also down, at 202.1 billion baht, 10.5% lower than the same period of the previous year. When exports are struggling, the value of handling more product types and pack formats with the same headcount, and of running the line without depending on the availability of workers, increases.

Finally, the global trend. According to World Robotics 2026, released by the International Federation of Robotics (IFR) on September 24, 2026, more than 600,000 new industrial robots were installed worldwide in 2025, up 11% year on year. The total number of robots operating in factories (operational stock) reached a record 5 million, up 9% year on year. By country, 2025 installations were 354,000 units in China (59% of the world), about 38,500 units in the United States (up 12% year on year), overtaking Japan for 2nd place, and 36,219 units in Japan (down 19% year on year). The IFR forecasts installations of 655,000 units in 2026 and 806,000 units in 2029 (both are forecasts). For reference, the previous edition, World Robotics 2025, reported 542,000 global installations in 2024, with 74% of new installations in Asia. Note that the World Robotics 2026 press release does not include figures for Thailand, ASEAN or the food industry.

Securing and paying for labor, export headwinds, and the global spread of automation: with these 3 factors overlapping, inquiries along the lines of “Could we put a robot on the tray-loading process?” are increasing. The first thing to check at that point is the nature of the parallel link robot as a robot type.

What Is a Parallel Link Robot? A Comparison with SCARA and Articulated Robots

A parallel link robot is a robot in which several arms (links) hang down in parallel from a base on the ceiling side and move a platform at their tips. Because its shape resembles the Greek letter Δ, it is also called a delta robot. Kawasaki Heavy Industries’ Y series catalog uses the term “Parallel link type” and describes the robots as long-reach, ultra-high-speed picking robots used in the food, pharmaceutical, cosmetics, electrical and electronics, and machine parts sectors (manufacturer’s description). The datasheet for OMRON’s iX4 describes a structure in which 4 arms distribute the load evenly and states that it supports high-speed pick and place on high-speed conveyors (manufacturer’s description).

For mounting, ceiling mounting (Ceiling, inverted) is the norm across manufacturers. The FANUC, ABB, Kawasaki Heavy Industries and OMRON documents all list ceiling mounting. Unlike floor-mounted robots, the robot works from directly above the conveyor, so it tends not to take up space on the line. On the other hand, the rigidity of the mounting frame and the cleanability of the frame itself become design issues.

There are variations in the number of axes. The mainstream is 4 axes, consisting of 3 translational axes plus 1 wrist rotation axis; examples include the FANUC M-2iA and DR-3iB, the 4-axis version of the ABB IRB 360, the Yaskawa MOTOMAN-MPP3S, the Kawasaki Heavy Industries YF002N/YF003N, and the OMRON iX4/Quattro. There are also 3-axis machines for applications that do not need rotation (such as the FANUC M-1iA/1H), 5-axis machines for re-orientation applications that change product orientation (ABB IRB 365), and some 6-axis variations (FANUC M-1iA).

The work envelope is expressed as “diameter × height”. For example, the Kawasaki Heavy Industries YF003N is Φ1,300×H500, and the OMRON iX4-650 has a stroke of 1300 mm in the X and Y directions and 500 mm in the Z direction. The work envelope is not a simple cylinder; the reachable range narrows at the outer edge and at the bottom, so checking the outline drawing is essential.

The differences from SCARA robots and vertical articulated robots can be summarized as general characteristics as follows. Because we could not find a primary source in which a manufacturer compared the robot types numerically, this parallel link robot comparison stays qualitative.

For an overview of price ranges and how to choose among robot types, see “Industrial Robot Implementation 2026: Prices and Selection of 5 Types Including Articulated and SCARA“. This article narrows the focus to the parallel link type and covers the practical work of model selection, the RFP and acceptance testing.

How to Read Catalog “Cycle Time” and “Picks per Minute”

The area where misunderstandings most often occur in delta robot selection is the speed figures in catalogs. In short, manufacturer-published values change considerably with test conditions, so you cannot line up figures across manufacturers and compare “which one is faster”.

“25-305-25” Is Shared, but the Content Is Not the Same

Many manufacturers use a test motion called “25-305-25”: a 25 mm rise, a 305 mm horizontal move, and a 25 mm descent. However, each company describes the conditions differently.

• Kawasaki Heavy Industries’ catalog describes it as a to-and-fro motion of “25 mm up, 305 mm horizontal, 25 mm down”.
• OMRON’s datasheet lists, in addition to 25/305/25, a 25/700/25 condition with the horizontal move extended to 700 mm, and states that the values are sustained cycle times at an ambient temperature of 20°C.
• ABB’s IRB 360 datasheet (2013 edition) shows 2 conditions, 25/305/25 and 90/400/90, and notes that the values were measured under real conditions but may vary depending on the actual application (tool data, path rounding radius, gripper actuation, etc.).
• Yaskawa’s MPP3S datasheet shows 3 patterns, 300-400-100 and 70-700-70 in addition to 25-305-25, and notes that the cycle times in the table are “without picking time and product”.

In other words, even under the same “25-305-25” name, the conditions are not aligned: round trip or one way, how many kg were loaded, whether pick and place time is included, and whether the value is sustained or momentary.

On the Same Model, the Figure Changes Depending on Whether Pick and Release Time Is Included

A good example of how much the conditions matter is Yaskawa’s MPP3S. The US manufacturer page publishes up to 150 CPM (cycles per minute) with a full 3 kg payload and up to 185 CPM with 1 kg, and states explicitly that these values include 0.04 seconds per cycle for part acquisition and release (manufacturer-published value). Meanwhile, the table in the datasheet for the same model lists CPM under the condition “without picking time”, and these figures are larger for the same payload. Even for the same model from the same manufacturer, the stated capacity looks quite different depending on whether pick and release time is included.

On a real line, there is always time for the suction cup to establish vacuum, time to break the vacuum and release the workpiece, and time to wait until the workpiece has reliably separated. Depending on the shape and surface condition of the workpiece, this can take longer than the catalog assumes.

On Some Models, Payload Changes Considerably with or without Rotation

Another pitfall is the payload. In the datasheet for the OMRON iX4-650H, the maximum payload is listed as “6 kg (P30: 15 kg)”. P30 is the maximum payload with a fixed platform without rotation; with a rotating platform the payload drops considerably. For the iX4-800H it is “4 kg (P30: 10 kg)”. Processes that align products in a set orientation often require a rotation axis, and in that case the usable payload is not the largest number in the catalog.

The payload calculation includes not only the workpiece weight but also the weight of the hand (suction cups, manifold, tubing, mounting bracket). Multi-pick hands that pick several items at once tend to be heavy.

From Catalog Values to Real-Line Capacity

To summarize, the gap between catalog values and real-line capacity arises from an accumulation of factors such as the following.

• Pick and release time (establishing vacuum, breaking vacuum, confirming separation)
• Actual travel distance and vertical stroke (the positional relationship between the conveyor and the tray, the depth of the tray pockets)
• Conveyor tracking and vision processing (picking while following moving products)
• Whether rotation is used and the angle (with rotation, payload and motion conditions change)
• Heat and load in continuous operation (OMRON explicitly states that its cycle times are sustained values)

Given this gap, the starting point for selection is not the catalog but “the capacity our process needs”. This article uses the following assumed example.

60 pieces/min per robot means, on average, 1 pick-and-place every 1 second. In the acceptance test, this is multiplied by a margin factor of 1.2, and the robots run continuously at 72 pieces/min per robot (60×1.2) and 144 pieces/min for 2 robots combined (120×1.2). The margin factor of 1.2 is a value proposed in this article, not an industry standard. Decide it yourselves based on the variation in flow rate, the size of peaks, and the need to catch up after stoppages.

Delta Robot Comparison: Published Specifications of Major Manufacturers’ Models

Here we list the specifications stated on official pages and datasheets for representative models from the major manufacturers. All values in the table are manufacturer-published values (manufacturer claims). As explained in the previous section, the test conditions are not aligned, so we do not compare speeds across manufacturers using cycle times or picks per minute. Speed is something to verify by having the manufacturer demonstrate with your own workpieces and hand.

Some notes on reading the table:

• The ABB IRB 360 values are from the 2013 edition datasheet (the IRC5 controller generation). They may differ from current specifications, so check the latest documents when requesting quotations.
• The Kawasaki Heavy Industries values are from the catalog (Cat. No. 3L2246 May ’22). The cycle times in the same catalog are 0.3 s (0.5 kg) / 0.36 s (2 kg) for the YF002N and 0.27 s (1 kg) / 0.45 s (3 kg) for the YF003N, under the to-and-fro motion described above (catalog values). Other documents show different values in some cases, so watch for differences between editions.
• The HS type of the OMRON iX4 is listed as USDA-Accepted for meat and poultry processing, and its platform is made of stainless steel. Its payload differs from the H type; the iX4-650HS is “3 kg (P30: 12 kg)”.
• Published repeatability is ±0.04 mm for the Kawasaki Heavy Industries YF002N (per ISO 9283), ±0.1 mm for the YF003N, ±0.1 mm for the Yaskawa MPP3S, ±0.10 mm for the OMRON iX4-650, and ±0.02 mm for the FANUC M-1iA/1H. However, in processes that pick products from a conveyor, what matters is the accuracy during tracking, discussed later, rather than static repeatability.
• Within the scope of this research, we could not confirm in primary sources that any major manufacturer announced a new delta robot model in 2025-2026. Judge models by whether they fit your process conditions, not by how recently they were announced.

This table is material for “narrowing down candidates”, not for “deciding” on a model. For example, if the weight including the hand becomes large with rotation, review the candidates using the payload with rotation rather than the P30 figure. If the robot has to cover a wide area including the conveyor width and tray positions, check on the outline drawing that the reachable range narrows at the outer edge of the work envelope. If washdown is a precondition, narrow down candidates using the hygienic specifications discussed in the following sections. Proceeding in this order and narrowing down to 2-3 models before requesting demonstrations is the realistic approach.

Hand and Workpiece: The Grip That Determines High-Speed Pick and Place Robot Capacity

What ultimately determines the capacity of a high-speed pick and place robot is the combination of hand and workpiece, more than the robot itself. Catalog cycle times are often measured with light test loads, and the conditions change the moment you attach your actual product and hand.

Vacuum or Gripping

In tray-loading processes with parallel link robots, vacuum suction is widely used, because it is light, picks and releases quickly, and allows a small hand. On the other hand, suction may not be stable with workpieces whose surfaces are uneven or porous, workpieces with oil or moisture on them, or soft workpieces that easily lose their shape. In such cases, consider gripping with fingers or a dedicated hand designed for the workpiece. The approach for each hand type and gripping force calculations are covered in detail in “Robot Hand and Gripper Selection 2026: 3 Types and Practical Gripping Force Calculation“, and how to choose a vacuum source for suction in “Vacuum Ejector Selection: Suction Handling Design and Acceptance Testing for Thai Plants“.

Measure Pick and Release Time

As the Yaskawa example in the previous section shows, pick and release time directly affects capacity. For a process at 60 pieces/min per robot, everything in 1 cycle, including movement, tracking, pick and release, must fit within 1 second on average. Because pick and release time varies with vacuum build-up, how the workpiece separates, tubing length and so on, measuring with actual workpieces and a prototype hand is more reliable than calculating on paper. Requesting “measured pick and release times with actual workpieces” at the RFP stage makes the later capacity calculation less likely to drift.

Multi-Pick as an Option

Multi-pick, which picks several items in 1 motion, is an effective way to increase capacity per robot. OMRON cites a high payload that supports multi-hand (multi-picking) operation as a design feature of the iX4 (manufacturer’s description). However, multi-pick hands are heavier and more likely to hit the payload limit with rotation. In addition, picking several items at once requires the spacing and orientation of products on the conveyor to be reasonably consistent. If you plan on multi-pick, check both the weight including the hand and the way products flow first.

Handling Drops and Misplacements

The faster the robot moves, the more likely workpieces are to fall due to centrifugal force and acceleration/deceleration. How dropped workpieces are handled (put back into the line, discarded, who picks them up) affects both hygiene management and safety. The allowable drop rate and misplacement rate are items that should be set numerically in advance as pass/fail criteria for the acceptance test.

Conveyor Tracking and Vision, and Load Balancing Across Multiple Robots

In tray-loading processes with parallel link robots, “conveyor tracking”, picking moving products without stopping them, is a precondition. The design here greatly affects real-line capacity and the number of missed picks.

Conveyor Tracking Basics

The conventional basic method is to attach an encoder to the conveyor to obtain speed and position, combine this with the product positions captured by a camera, and have the robot follow the movement of the products. Each manufacturer offers dedicated software and functions.

• FANUC’s iRPickTool is described as a software package that simplifies the setup and operation of line tracking and can handle multiple conveyors, fixed stations, trays, and load balancing among multiple robots. The FANUC Europe page shows values such as up to 32 robots, up to 32 conveyors, and up to 2000 picks/min (with multiple robots cooperating) (all manufacturer-published values).
• FANUC’s Real-time Visual Tracking performs line tracking with a high-speed 2D camera without using an encoder, and is intended for situations where mounting an encoder is not practical (manufacturer’s description).
• FANUC also provides a vision process for line tracking with iRVision 3D vision.
• ABB states that PickMaster can follow high-speed conveyors.
• OMRON states that the iX4/Quattro has 4 conveyor tracking inputs and that up to 8 robots can be connected via EtherCAT with the NJ501-R controller (manufacturer-published value).
• Yaskawa’s MPP3S supports MotoSight, MotoPick and conveyor tracking, and Yaskawa advises contacting its sales office when planning a multi-robot system.

Ordering and acceptance of conveyor tracking with 3D vision are covered in detail in “3D Vision Conveyor Tracking Implementation: An Ordering and Acceptance Guide for Thai Plants“.

Tracking Accuracy Drops as Conveyor Speed Increases

The relationship between conveyor speed and tracking accuracy is easily overlooked. ABB’s IRB 360 datasheet (2013 edition) lists the repeatability during tracking measured under real conditions with the IRB 360-1/1130 and PickMaster: 1.0 mm at a constant conveyor speed of 200 mm/s, 1.5 mm at 350-750 mm/s, and 5.0 mm at 800-1400 mm/s (manufacturer-published values). Static position repeatability in the same datasheet is 0.1 mm, which shows that the faster the conveyor, the larger the deviation in pick position.

In processes that place products into tray pockets, this deviation determines whether the product fits into the pocket. If you keep the product flow rate the same, reducing the spacing between products can sometimes be more advantageous than speeding up the conveyor. It is important to decide conveyor speed, product spacing and tray tolerances as a set.

What to Decide for Vision

The role of vision is to detect the position and orientation of products and, if necessary, to identify products that must not be picked (wrong orientation, overlapping, chipped). The points to decide are as follows.

• What to detect (only position and orientation, or appearance inspection as well)
• Lighting and background (conveyor belt color, product gloss, influence of ambient light)
• How to handle products that overlap or touch (let them pass without picking, or align them upstream)
• The distance from the imaging position to the robot’s work envelope, and product shifts in between

For how to think about the cost and benefit of vision, see also “Robot Vision Implementation Cost and Payback: Determined by Feeding, Accuracy and Lighting“.

Load Balancing with 2 Robots

In this article’s example, the flow of 120 pieces/min is shared by 2 robots, at 60 pieces/min per robot. A common configuration is for the downstream robot to pick the products the upstream robot could not pick.

The points to decide for load balancing are as follows.

• Sharing rule: Do the upstream and downstream robots each pick half, or does the upstream robot pick as many as it can and the downstream robot pick the rest? The latter tends to make the downstream load more variable.
• Where missed picks go: How products that neither robot could pick are collected at the end of the conveyor.
• Operation when 1 robot stops: When 1 robot stops, how many pieces/min the other robot can handle, and whether to slow the line or stop upstream. For some candidates we could not confirm in manufacturers’ primary sources whether there is a function that automatically redistributes the load, so check this as a specification for each candidate model and actually test it in the FAT.
• Tray-side synchronization: If the trays are also moving on a conveyor, tray-side tracking is also needed.

Hygienic Design and Washdown for Food and Pharma (Washdown Food Robot, IP69K)

When parallel link robots are used in food or pharmaceutical plants, hygienic specifications strongly influence model selection. For a robot that moves directly above the product, lubricant leaks and structures where dirt easily accumulates translate directly into a risk of foreign matter contamination.

Manufacturers’ Hygienic Specifications (Published Values)

The hygiene-related specifications stated in the major manufacturers’ documents are as follows (all are manufacturer-published values and descriptions).

IP69K is often referred to in discussions of food plant washdown specifications as a protection rating that assumes cleaning with high-pressure, high-temperature water. However, choosing an IP69K model does not complete the hygienic design. You need to check whether the robot itself, the hand, tubing, cables and frame can each withstand the type of detergent or disinfectant used (acid or alkaline), the temperature, the pressure and the frequency.

“Food Grade” and NSF H1

In food plants, a lubricant category called NSF H1 often comes up. NSF H1 generally refers to the registration category for lubricants used where incidental food contact may occur. In the manufacturers’ primary sources we checked, all used the wording “food grade grease”, and no specific registration category was stated. If your own quality standards or customer audits require a specific registration category, confirm with the manufacturer in writing the name of the lubricant used and whether it is registered.

EHEDG Guideline 62 Part 1

As a hygienic design standard, EHEDG (European Hygienic Engineering & Design Group) published Guideline 62 Part 1 as a first edition in March 2026. Its title is “Hygienic Design Criteria for Robotic Systems in Food Processing Environments Part 1: Fundamental Requirements for Robots”. Its scope is open food processing environments; it covers both wet and dry cleaning and the minimum requirements for construction, installation, integration and documentation. It is also used as the basis for assessing hygienic design in EHEDG’s equipment certification program. Future parts are planned to cover end effectors (hands), system integration, mobile robots and other topics.

The hygienic design of hands and of the system as a whole is an area planned to be covered in future parts. For now, the hygienic design of the robot itself can refer to the guideline, while the hand, frame, tubing and covers need to be worked out using your own hygiene standards and the designs of the manufacturer and SIer (system integrator).

Decide the Cleaning Procedure First

In practice, the most effective step in selecting hygienic specifications is deciding the cleaning procedure first. Will you wash with high-pressure hot water at the end of each day, or is dry cleaning enough? Will the hand be removed when the robot is washed? Will wash water pool on top of the frame or in cable ducts? Once these procedures are decided, the required protection rating, materials and whether covers are needed follow naturally. Conversely, if you select a model only by protection rating without deciding the procedure, you may find that the hand or frame cannot withstand washdown.

Safety Design for Delta Robots (Guarding, ISO 10218-1/-2:2025, ISO 13857)

A parallel link robot is a high-speed industrial robot, not a collaborative robot. This article does not assume that people work in the same space without fencing; it assumes the robot is enclosed by guarding. The Kawasaki Heavy Industries catalog also notes that many of its photos were taken with the safety fences and safety devices required by safety regulations removed, which assumes that safety fences and similar measures are needed on actual machines.

The ISO 10218:2025 Revision

The safety standard for industrial robots, ISO 10218, was revised in February 2025. ISO 10218-1:2025 is the 3rd edition and covers safety requirements for the robot itself, and ISO 10218-2:2025 is the 2nd edition and covers safety requirements for robot applications and robot cells. According to a trade journal’s explanation, the main changes include making functional safety requirements explicit rather than implicit, integrating the content of the collaborative operation technical specification ISO/TS 15066, incorporating safety guidance on end effectors and manual loading/unloading, updating robot classifications, and adding cybersecurity requirements. It is described as the largest revision since the 2011 edition.

When introducing a parallel link robot, the division of responsibility is that the manufacturer is responsible for the robot’s specifications, and the SIer and user are responsible for the safety design of the cell as a whole. It is important to state at the RFP stage which edition of which standard the risk assessment and safety design will be based on.

ISO 13857 Safety Distances

The standard for deciding guard height and opening size is ISO 13857:2019. It defines safety distances to prevent upper and lower limbs from reaching danger zones; the 2nd edition published in 2019 was reviewed and confirmed in 2025. Human body dimensions assume the 95th percentile. Because the conveyor entry and exit require openings in the guarding, use this standard to check the distance at which hands cannot reach through the openings.

Guarding Configuration

Generally, maintenance doors are fitted with interlocks so that the robot stops when a door is opened. Because the conveyor entry and exit are openings through which products pass, measures such as securing the ISO 13857 safety distance or fitting tunnel-shaped covers are taken. ABB’s IRB 360 datasheet lists double circuits with supervision, emergency stops, safety functions and a 3-position enabling device, and shows collision detection as an option (manufacturer’s description).

In food plants, the guarding itself is also subject to cleaning. Hygiene checks are also needed, such as whether the transparent panel material can withstand detergents and whether dirt accumulates at the bottom of the guarding. How to decide whether safety fences are needed and how to think about stopping distance are explained in detail in “Robot Safety Fence Standards 2026: Stopping Distance, Not the Fence, Decides Whether You Need One“.

Delta Robot Cost and an Illustrative Cost-Benefit Model (Assumed Figures)

Regarding delta robot cost, none of the official pages or datasheets from FANUC, ABB, Yaskawa, Kawasaki Heavy Industries or OMRON that we checked showed public prices. Actual costs vary considerably depending on the scope of vision, hand, conveyor, safety fencing (guarding), hygienic specifications and SI (system integration), in addition to the robot itself and the controller. Be sure to confirm through quotations.

With that in mind, we show an approach for discussing cost-effectiveness internally using an illustrative model calculation with assumed figures. All of the figures below are assumed values set up for this article.

Model Plant Assumptions

The scope is only the same single line as in the example in the previous sections, not the entire plant. The currency is Thai baht (THB), and the line runs 3 shifts. There is only one baseline for comparison: “current state = manual tray loading by 12 people”.

Under these assumptions, the annual net benefit is 1,320,000 THB and the simple payback period is 4.0 years.

How the Payback Period Moves When the Headcount Reduction Changes

The headcount reduction is the input in this calculation that depends most on assumptions. Keeping the initial cost and operating cost fixed and changing only the headcount reduction gives the following.

If the headcount reduction falls from 9 to 6 people, the payback period extends from 4.0 years to 6.8 years. The headcount reduction is not “a value a robot is guaranteed to achieve”; it is an assumed figure meant to show how much the benefit depends on the assumptions. Even after automation, people are still needed for product feeding, tray replenishment, collecting missed picks, changeovers and cleaning. Replace these values with actual figures from an on-site work analysis and the FAT results.

What the Calculation Does Not Include

• BOI tax incentives are not included in the payback calculation. If the project qualifies for the BOI efficiency improvement measure described later, corporate income tax exemption and other incentives may be available, but eligibility needs to be confirmed case by case.
• Quality improvement and yield improvement effects are not counted either. Reducing misplacements and foreign matter contamination risk in manual tray loading may have qualitative benefits. However, to keep a single baseline and avoid double counting benefits, the calculation here uses labor cost reduction only. In internal approval documents, it is safer to describe such effects separately as “upside factors not included in the figures”.

What to Write in a Delta Robot RFP

The issues so far are organized below as items to write in the RFP. The key point is to request not “presentation of catalog values” but “a demonstration under our conditions and commitments under those conditions”.

Items 6 and 11 are especially important. Because each company’s published values cannot be compared on the same footing, have each supplier state “under which conditions the figure was obtained”, and then have them demonstrate under your conditions. Combining these two makes proposals easier to compare.

What to Confirm in FAT/SAT

The FAT is the acceptance test conducted before shipment in an environment such as the SIer’s facility, and the SAT is the acceptance test conducted after installation at your own plant. The core of FAT/SAT for parallel link robots is to run continuously at the target capacity multiplied by the margin factor and confirm that missed picks, drops and misplacements are within the allowable range.

In the FAT, capacity and functions are confirmed in the SIer’s environment; in the SAT, the same items are rechecked with your own products, your own conveyor and your own cleaning procedure. Washdown and hygiene in particular can only be verified in your own environment. The actual capacity obtained in the FAT and the reality of the work remaining after automation also serve as material for replacing the headcount reduction in the model calculation in the previous section.

For how to think about the whole process including case packing, see also “Case Packing Automation Robots 2026: It Is Not the Packing Step That Stops the Line“.

Issues Specific to Thailand and ASEAN

1. The BOI Efficiency Improvement Measure

The Thailand Board of Investment (BOI) has an efficiency improvement measure targeting machinery replacement and automation in existing businesses. It is the “Measure to improve the efficiency in upgrading or replacing of machinery and automation” within the “Measure for Industrial Upgrades towards Smart and Sustainable Industry” under BOI Announcement No.15/2565 (December 8, 2022). According to the BOI investment promotion guide, the main points are as follows.

• Eligibility: Existing businesses (whether or not BOI-promoted). Investment of 1 million baht or more (excluding land and working capital). For BOI-promoted projects, the corporate income tax exemption period must have ended
• Incentives: Exemption of import duties on machinery, and a 3-year corporate income tax exemption (capped at 50% of the investment)
• For automation and robot investments where procurement from the Thai automation industry (linkage) is 30% or more of the value of the machinery being improved, the cap is 100% of the investment
• Operations must start within 3 years of the issuance of the promotion certificate

The application deadline could not be determined from the documents we checked. In addition, media reports in July 2025 described expanded measures for Thai small and medium-sized enterprises (SMEs), but these are said to target Thai SMEs and are likely not applicable to large Japanese-affiliated companies. Check individually with the BOI or a specialist whether your company is eligible and which measures you can use. As mentioned above, the model calculation in this article excludes BOI incentives.

2. How to View the Minimum Wage and Labor Costs

As mentioned above, the minimum wage differs by region, and the effective date may differ depending on the notification. When estimating cost-effectiveness, thinking in terms of the total cost per person, including social security, bonuses, overtime, and the effort of hiring and training, rather than the minimum wage, gets you closer to reality.

3. Local SIers and Spare Parts

A parallel link robot only works once vision, tracking, the hand, guarding and hygienic specifications are combined with the robot itself. When introducing one in Thailand, check whether there is a local SIer that can design, install and commission all of these together, whether failures can be handled locally, and whether consumables such as suction cups and spare parts can be procured within Thailand. High-speed robots also tend to require more frequent replacement of hand consumables. Deciding the spare parts list and how to hold inventory at the RFP stage helps you avoid trouble after start-up.

4. Language for Operation and Training

Local staff handle changeovers and recovery after stops. We recommend including the language of operating screens and procedure manuals, and the scope and number of training sessions, among the items for comparing proposals.

Frequently Asked Questions

What is a parallel link robot?

It is a robot in which several arms hang down in parallel from a base on the ceiling side and move a platform at their tips; it is also called a delta robot. It is ceiling-mounted and excels at pick and place that moves light workpieces horizontally at high speed. It is often used in the food, pharmaceutical, cosmetics, electrical and electronics sectors for processes that pick products from a conveyor and arrange them in trays or cases.

How much does a delta robot cost?

The major manufacturers’ official pages and datasheets do not show public prices. Costs vary considerably depending on the scope of vision, hand, conveyor, guarding, hygienic specifications and SI, in addition to the robot itself and the controller, so they need to be confirmed through quotations. This article shows an illustrative calculation with assumed figures for internal discussion, but these are not actual amounts.

How should delta robots be compared?

Rather than lining up catalog cycle times or picks per minute across manufacturers, narrow down candidates by payload (with or without rotation), work envelope, hygienic specifications and tracking functions, and compare capacity through demonstrations with your own products and hand. This is because test conditions for published values differ by manufacturer, and even on the same model the figures change depending on whether pick and release time is included.

Should I choose a delta robot or a SCARA robot?

Generally, parallel link robots suit processes that rearrange light workpieces on a conveyor at high speed, while SCARA robots suit floor-mounted processes such as part insertion and alignment. However, the answer changes depending on workpiece weight, required orientation and installation space. For an overview of each robot type, refer to the related article.

Can a parallel link robot be washed down with water?

It depends on the model. According to manufacturer-published values, some models support IP69K and some support washdown with acid or alkaline cleansers. However, you need to check whether not only the robot itself but also the hand, tubing, cables, frame and guarding can withstand washdown. Deciding the cleaning procedure (detergent, temperature, pressure, frequency) before selecting a model is the reliable approach.

What is needed for conveyor tracking?

Generally, you need an encoder to capture conveyor speed and position, a camera to detect product position and orientation, and a tracking function on the robot side. Manufacturers also offer methods that track with a high-speed camera without an encoder. Because tracking accuracy tends to decrease as conveyor speed increases, speed, product spacing and tray tolerances are decided as a set.

How do I decide whether 1 robot is enough or 2 are needed?

Divide the required flow rate by the capacity that 1 robot can actually achieve with your products and hand, and allow for a margin factor. In this article’s assumed example, 120 pieces/min is shared by 2 robots, and the design confirms in the acceptance test whether each robot can run continuously at 72 pieces/min. How the line operates when 1 robot stops is also a factor in deciding the number of robots.

Conclusion

• Delta robot selection is not decided by the catalog’s maximum speed or maximum payload. Test conditions for published values differ by manufacturer, and even on the same model the figures change depending on whether pick and release time is included.
• Real-line capacity is determined by pick time with actual workpieces and an actual hand, conveyor tracking and vision, whether rotation is used, and washdown and hygiene conditions. On some models, the payload drops considerably with rotation.
• Proceed with selection in this order: (1) define the required capacity, (2) workpiece and hand, (3) narrow down models, (4) tracking and load balancing, (5) safety, hygiene and acceptance.
• In food and pharma, decide the cleaning procedure first, and check hygienic specifications not only for the robot itself but also for the hand, frame and guarding. Confirm the lubricant category in writing.
• Cost-effectiveness depends heavily on the headcount reduction assumption. In the illustrative calculation with assumed figures, the payback was 4.0 years with a reduction of 9 people and 6.8 years with 6 people. Replace these values with actual figures from a work analysis and the FAT.
• In the RFP, request a demonstration under your conditions; in FAT/SAT, run continuously at the target capacity multiplied by a margin factor and confirm that missed picks, drops and misplacements are within the allowable range.

TOMAS TECH helps plants in Thailand automate robotic tray-loading and picking processes to suit their lines and products. You are welcome to consult us even before a model has been decided, for example while you are still organizing the flow rate of the target process or narrowing down candidate models. Please feel free to contact us through this form.

References

• FANUC: Delta robot lineup
• FANUC America: M-1iA/1H
• FANUC America: M-1iA/0.5S
• FANUC America: DR-3iB/8L
• FANUC America: Announcement of a food-grade delta robot (September 23, 2019)
• FANUC America: iRPickTool
• FANUC Europe: iRPickTool Advanced Line Tracking
• FANUC America: Real-time Visual Tracking
• FANUC America: iRPickTool iRVision 3D
• ABB: IRB 360 datasheet (May 2013)
• ABB: IRB 365 launch announcement
• Yaskawa America: MPP3S
• Yaskawa Europe: MPP3S datasheet
• Kawasaki Heavy Industries: Y series catalog
• OMRON: iX4 / Quattro datasheets
• ISO: ISO 10218-1:2025
• ISO: ISO 10218-2:2025
• ISO: ISO 13857:2019
• Control Design: Explanation of the ISO 10218 revision
• EHEDG: Publication of Guideline 62 Part 1 (March 24, 2026)
• EHEDG: Guideline 62 Part 1
• IFR: World Robotics 2026 announcement (September 24, 2026)
• IFR: World Robotics 2025 announcement
• BOI: Smart and Sustainable Industry measure
• Nation Thailand: BOI measures for SMEs (July 14, 2025)
• Thailand Ministry of Labour: Minimum wage revision (July 17, 2025)
• Thairath: Minimum wage in 2026 (for reference)
• Nation Thailand: Thai food exports (March 31, 2026)