Collaborative Robot Implementation 2026: Thailand Safety, TCO & RFP Guide
Collaborative robot implementation in a Thai factory should not begin with a robot price list. It should begin with evidence that the complete application—robot, end effector, workpiece, surrounding machinery, people, teaching and maintenance—can meet the required cycle time safely and economically. A cobot is not a fence-free turnkey machine by default. This 2026 buyer guide turns that principle into an actionable process for application screening, risk assessment, TCO comparison, RFP preparation, FAT/SAT and a gated 90-day pilot.
Executive answer: buy a viable production application, not a robot arm
The most reliable sequence is to define the process first, set technical, safety and business acceptance criteria, and only then compare robots and integrators. Before issuing an RFP, production, engineering, EHS, maintenance, quality, IT and procurement should agree on seven questions:
- What exactly must be moved, processed, assembled or inspected?
- What cycle time, operating hours, changeovers and product mix are required?
- Does a person genuinely need to share the workspace with the robot?
- What are the workpiece weight, centre of gravity, surface, tolerances and hazardous features?
- What happens during loading, teaching, recovery, cleaning and maintenance—not only normal operation?
- Which baseline measures will prove improvement in labour, quality, downtime and safety?
- Who decides whether the pilot scales, changes or stops?
If these questions remain open, comparing payload, reach and arm price usually creates late additions: guarding, fixtures, vision, machine modifications, controls integration, training and support. For partner due diligence, use our Thailand robot system integrator selection guide. For the investment model, see the robot implementation ROI guide.
A collaborative robot does not automatically mean “no fence”
A collaborative robot provides functions that can support applications where people and robots share space under defined conditions. The product category alone does not make every application safe. A slow-moving arm can still present unacceptable risk when it carries a sharp tool, hot part or heavy load; drops a workpiece; traps a person against a fixed structure; or triggers surrounding machinery.
Application safety depends on the combined system:
- robot and safety-related functions;
- gripper, process tool, cables and hoses;
- workpiece shape, mass, temperature, surface and centre of gravity;
- fixtures, stands, conveyors and process machinery;
- layout, access routes and escape space;
- speed, force, stopping behaviour and restart logic;
- normal operation, setup, teaching, recovery, cleaning and maintenance;
- user competence, authority and foreseeable misuse.
Whether fencing is needed cannot be decided from one brochure statement. A risk reduction strategy may support an open application, or it may require partial guards, scanners, interlocked access and monitored speeds. If simultaneous human access creates no process value, physical separation with faster robot motion may produce a better safety and throughput outcome.
How to use ISO 10218:2025 and ISO/TS 15066 in procurement
ISO published ISO 10218-1:2025 and ISO 10218-2:2025 in February 2025. Part 1 addresses industrial robot design; Part 2 addresses robot applications, cells and integration. The 2011 editions are withdrawn. A 2026 RFP should therefore identify the edition, scope, responsible party and required evidence rather than say only “ISO 10218 compliant.”
ISO/TS 15066:2016 provides additional guidance for collaborative operation and is under revision. Do not describe it as superseded unless an official source confirms that status. Applicable standards, Thai legal obligations, customer standards and corporate EHS requirements must be confirmed for each project.
A current Universal Robots safety manual states that the application integrator—which may be the user—is responsible for the risk assessment using ISO 12100 and ISO 10218-2, with ISO/TS 15066 as additional guidance. The assessment should consider normal operation, teaching, troubleshooting, maintenance, collisions, sharp end effectors, dropped workpieces and changes to safety settings.
| Required deliverable | What to verify | Typical participants |
|---|---|---|
| Applicable standards list | Edition, scope, exclusions and customer requirements | Integrator, EHS |
| Risk assessment | Lifecycle hazards, initial risk, measures and residual risk | Integrator, user |
| Safety requirements specification | Functions, triggers, stop/restart and access rights | Integrator, engineering |
| Verification records | Test conditions, instruments, results and deviations | Integrator, quality, EHS |
| Operating and maintenance information | Setup, recovery, inspection and change control | Integrator, maintenance |
| Training records | Role-based competence and retraining triggers | User, integrator |
Standards are a framework, not a substitute for project-specific professional judgment. Complete the application risk assessment and obtain qualified safety advice where required.
Screen candidate applications before selecting a cobot
In the first two weeks, walk the process and compare opportunities with the same scorecard. Start with a measurable problem—ergonomic strain, unstable quality, scarce labour, repetitive handling or known downtime—not a vague objective to “replace an operator.”
Stronger first-pilot candidates
- machine tending where most motions and part presentation can be standardised;
- repetitive fastening, dispensing or light processing along a controlled path;
- loading a measuring device, recording results and sorting parts;
- packing, case loading or lightweight palletising;
- human judgment combined with robot repetition;
- tasks where automation reduces ergonomic or repetitive exposure.
Higher-complexity first-pilot candidates
- random presentation or unstable part orientation without a feeding strategy;
- virtually no cycle-time margin and frequent waiting for human decisions;
- sharp, hot, fragile or heavy parts with severe drop or pinch consequences;
- many upstream and downstream stoppages outside the cell’s control;
- acceptance criteria based only on undocumented operator judgment;
- frequent manual intervention without a defined safe recovery procedure.
These applications are not impossible. They usually add vision, feeding, special tooling, guarding and control integration, making a first project’s cause-and-effect harder to isolate.
| Screening factor | Strong candidate | Weak candidate |
|---|---|---|
| Repeatability | Stable sequence, parts and criteria | Sequence and judgment change constantly |
| Cycle margin | Measured cycle has adequate margin | Even the ideal cycle is too tight |
| Part presentation | Position and orientation are controllable | Tangled, piled or frequently missing |
| Risk reduction | Hazards can be removed, separated or controlled | Sharp, hot or crushing hazards remain |
| Quality measurement | Acceptance can be measured | Only tacit judgment exists |
| Baseline data | Before/after data use the same definition | No reliable baseline |
| Product continuity | Demand and product life support the investment | End-of-life or redesign is near |
| Ownership | Production and users participate | Engineering works alone |
Use the score to structure discussion, not to average away an unacceptable safety condition. Safety-critical gates remain pass/fail.
Validate cycle time with real parts and abnormal conditions
Do not infer production capacity from maximum robot speed or an ideal path. A real cycle includes part detection, gripping, grip confirmation, travel, positioning, processing, release, machine communication, speed reduction when people approach, retries and recovery.
Actual cycle time = robot motion + process waiting + end-effector action and confirmation + communication + safety-related slowdown + average stoppage loss
Measure that equation with representative parts and a realistic layout. Study the distribution across variants, retries, shortages and changeovers—not only the fastest cycle.
An illustrative capacity check is:
Maximum allowable cycle = net available production time ÷ required good units
Define net available time consistently for the existing and proposed process. Any trial values must be labelled as assumptions, not supplier guarantees.
Tests should include the lightest and heaviest parts, extreme dimensions, position errors, missing and double feeds, grip failures, restart after an emergency stop, communication loss, maximum-reach poses and cable tension. Require input conditions, logs and repeatable procedures, not only a demonstration video.
Collaborative robot safety across the lifecycle
Risk assessment starts during concept design, not immediately before commissioning. Late safety additions can change speed, layout and ROI.
Normal production
Observe where hands, head and torso enter during loading and unloading. Examine trapping points involving the gripper, workpiece, fixture, table, column and adjacent machine—not only contact with the arm.
Teaching and changeover
People stand closer to the robot and in different postures during teaching. Define operating modes, permissions, enabling controls, speed limits and coordination between multiple workers. If recipe changes affect safety parameters, specify who may edit, approve and restore them.
Fault recovery
“Reach in and clear it quickly” is often the least controlled activity. Define fault types, energy isolation, manual controls, restart checks and supervision for every shift.
Cleaning, maintenance and modification
Include tool replacement, sensor cleaning, lubrication, wiring, software updates, backup restore and relocation. A new product or end effector may invalidate parts of the original assessment and trigger revalidation.
Residual risk must be connected to procedures, training, inspection, permissions, supervision and PPE. “The operator must be careful” should never replace feasible inherently safe design, guards or safety functions.
End-effector selection is a core engineering decision
The end effector is where the robot meets the process and often where failures originate. Selection should cover part condition, fault behaviour, sensing, cleaning and changeover—not just nominal grip force.
| End-effector type | Typical fit | Key checks |
|---|---|---|
| Parallel or multi-finger gripper | Stable parts, internal or external gripping | Tolerance, surface damage, pinch points |
| Vacuum gripper | Sheets, cartons and suitable bags | Leakage, dust, loss of vacuum, part detection |
| Magnetic gripper | Ferromagnetic parts | Material, residual magnetism, power-loss behaviour |
| Dedicated or combination tool | Product families and multiple tasks | Weight, change time, wrong-tool prevention, spares |
| Process tool | Fastening, sanding or dispensing | Reaction force, vibration, wear, process traceability |
Payload is not only the workpiece. Include the gripper, adapter, sensors, cables and hoses, plus centre of gravity and pose. Avoid approving a design at the edge of rated payload from drawings alone; validate it with manufacturer data, calculation tools and physical tests.
Define what happens during pneumatic pressure loss, vacuum leak, power loss and communication failure. Risk controls may include redundant confirmation, mechanical retention or preventing access below a suspended part.
Compare collaborative robot price through TCO
This guide does not invent a market price. Collaborative robot project cost varies with payload, reach, end effector, fixtures, vision, safety, controls, site conditions, validation, warranty and support. Ask every bidder to price the same scope and identify exclusions.
Project TCO = equipment, design and integration + factory preparation + safety validation + training and ramp-up + maintenance, consumables, licences and downtime + future changes + retirement costs − recoverable value
| TCO area | Examples | Procurement question |
|---|---|---|
| Robot package | Controller, pendant, standard cables | Are required options and safety functions included? |
| End effector | Gripper, vacuum, tool, sensors | Are all parts, spares and wear items included? |
| Fixtures and stand | Base, locating and quick-change fixtures | Who owns floor work, anchors and transport? |
| Sensing | Vision, force, code and presence detection | Are lighting, calibration and licences included? |
| Safety | Guards, scanners, interlocks and safety PLC | Are assessment, validation and records included? |
| Integration | PLC, MES, ERP and process equipment | Are interfaces, cybersecurity and boundaries defined? |
| Commissioning | Design, build, install, FAT and SAT | What causes extra travel, overtime or retesting cost? |
| Operation | Service, calibration, consumables and updates | What are annual cost, response and part lead time? |
| Change | New products, tools, layout and reassessment | Are source files, backups and revalidation available? |
For comparison, a simple illustrative payback formula is:
Simple payback period = initial investment ÷ annual net benefit
Annual net benefit must represent benefits that can actually be realised—such as redeployment, overtime reduction, additional good output or avoided quality loss—less added operating cost. Test pessimistic, base and optimistic scenarios for demand, uptime, maintenance and product life. Safety measures are not a lever to make payback look shorter.
BOI incentives: potential upside, not an assumed return
Thailand BOI’s current Smart and Sustainable Industry measure states a minimum efficiency-enhancement investment of THB 1 million, import-duty exemption on machinery, and a three-year corporate income tax exemption capped at 50% of eligible investment. Where domestic automation linkage reaches at least 30%, the cap may be 100%.
Eligibility is project-specific. Confirm the current activity, timing, eligible investment, domestic linkage evidence and tax position with BOI and tax advisers before placing orders. Do not build an unconfirmed incentive into the base ROI.
BOI reported 61 efficiency and sustainability applications totalling THB 7.071 billion in Q1 2026, many involving energy saving, machinery upgrades, automation/robotics and digital technology. This does not mean all 61 were cobot projects, nor does it prove any individual project’s ROI.
Build an RFP around evidence and responsibility
A robot implementation support RFP should define acceptance and interfaces, not merely request a product list. Provide every bidder with the current process breakdown, cycle distribution, representative parts, drawings, layout, utilities, machine interfaces, shifts, product mix, EHS rules, cybersecurity requirements and FAT/SAT conditions.
| RFP area | Required response | What to compare |
|---|---|---|
| Concept | Layout, workflow, capacity model and alternatives | Does shared human access add value? |
| Safety | Standards, assessment process, residual risk, deliverables | Is the complete application covered? |
| Cycle | Time breakdown, assumptions, margin and proof method | Is it more than an ideal robot cycle? |
| Quality | Detection, measurement, reject handling and history | Are acceptance and traceability clear? |
| Availability | Failure modes, spares, recovery and support | Are Thailand response and lead times credible? |
| IT/OT | Connectivity, users, logs and backup | Are access, ownership and end-of-support defined? |
| Change | New products, source, licences and revalidation | What can the factory maintain independently? |
| Commercial | Scope, exclusions, gates, warranty and delay | Who pays after FAT/SAT failure? |
Certifications and installation counts are useful but insufficient. Ask for relevant applications, local support, safety documentation quality and long-term change history. Where possible, speak with customers who have operated comparable cells for more than the launch period.
Define FAT and SAT before the purchase order
FAT and SAT criteria belong in the RFP and contract. “Works correctly” is not measurable.
At FAT, verify drawings, bill of materials, software versions and backups; run representative and boundary parts; measure cycle and quality; test missing parts, grip failure, sensor faults and communication loss; test safety functions, stop/restart and modes; and close or formally transfer every open item.
At SAT, verify the real floor, utilities, lighting, network and surrounding machines; use production parts and intended operators across relevant shifts; test end-to-end capacity, quality, data and recovery; execute setup, cleaning, maintenance and backup restoration; complete site risk assessment and safety validation; and confirm role-based training and handover documents.
Contractually define sample size, test duration, allowable stoppage, acceptance quality, cycle statistics and retest rules. Safety criteria should be mandatory individual gates, not points that can be offset by a high overall score.
A gated 90-day collaborative robot pilot
Days 0–30: Screen & Define
- Observe work, waiting, faults and quality decisions.
- Record baseline cycle distribution, labour, defects, downtime and rework.
- Confirm representative parts, boundary cases, demand and product life.
- Map human/robot roles, lifecycle tasks and foreseeable misuse.
- Start the concept risk assessment and compare layouts.
- Define KPIs, stop conditions, data owners and the gate committee.
Gate 1: Is automation appropriate, is collaboration necessary, and are success conditions measurable? If not, improve feeding, fixtures or standard work first.
Days 31–60: Prove & De-risk
- Test gripping, poses and cycle time with representative parts.
- Test errors, missing parts, grip failures and recovery—not only normal flow.
- Design risk reduction and assess residual risk.
- Include operators, maintenance, quality and IT in reviews.
- Finalise TCO scope, support requirements, RFP and FAT/SAT criteria.
- Treat BOI eligibility as an open verification item until confirmed.
Gate 2: Is there evidence that cycle and quality can be met with an acceptable risk reduction concept and budget range? If not, redesign the process, feeding, tooling or layout.
Days 61–90: Pilot & Decide
- Run the agreed products and shifts in a controlled pilot.
- Record good output, stoppage, retries, interventions and maintenance consistently.
- Have site personnel execute changeover, cleaning, teaching and recovery.
- Verify safety, training, documentation, backups and change control.
- Update TCO, benefit and sensitivity analysis with measured results.
- Decide to scale, continue in a limited scope, redesign or stop.
Gate 3: Can the intended users operate the application repeatedly, safely and maintainably under defined conditions? A stop or redesign decision is a valid pilot outcome if it prevents a larger production loss.
Final checklist before collaborative robot implementation
Process and capacity
- [ ] The current cycle is measured as a distribution, not only an average.
- [ ] Representative parts, boundary cases and future variants are defined.
- [ ] Waiting, communication, intervention and recovery are included in capacity.
- [ ] The business need for a collaborative mode can be explained.
Safety and standards
- [ ] The risk assessment covers the complete application.
- [ ] Applicable editions of ISO 10218, ISO 12100, ISO/TS 15066 and other requirements are confirmed.
- [ ] Normal operation, teaching, recovery, cleaning, maintenance and change are assessed.
- [ ] Residual risks, training, inspection, permissions and change control are handover conditions.
Technology and operation
- [ ] Payload is verified with the end effector, workpiece, centre of gravity, cables and hoses.
- [ ] Grip failure, power loss, communication loss and position error are test cases.
- [ ] PLC/MES interfaces, logs, backups and responsibility boundaries are defined.
- [ ] Thailand-based parts supply, response time and maintenance support are verified.
Commercial and investment
- [ ] Every bidder receives the same RFP and acceptance criteria.
- [ ] Proposals are compared through TCO and sensitivity analysis, not initial price alone.
- [ ] Unconfirmed BOI incentives are excluded from the committed base case.
- [ ] FAT/SAT, payment gates and remedies for failure are included in the contract.
FAQ
Does a collaborative robot always operate without fencing?
No. The complete application requires risk assessment. Depending on the end effector, workpiece, layout, speed and lifecycle tasks, an open design may be justified or guards, scanners, interlocks and separation may be required.
How much does collaborative robot implementation cost?
There is no reliable project price without scope. Compare the same TCO boundary: robot, tooling, fixtures, sensing, safety, controls, engineering, installation, validation, training, service, licences, consumables and change. This article intentionally avoids fabricated market prices.
Who is responsible for collaborative application safety?
Roles depend on contracts and applicable law, but responsibility cannot simply be delegated to the robot manufacturer. The integrator and user must define responsibility for application design, risk assessment, validation, training, operation and change. The user may itself act as the integrator.
What matters most when selecting a robot gripper?
Evaluate workpiece mass, centre of gravity, tolerances, surface, extreme poses, failure retention, sensing, cables, cleaning, wear and product change—not only nominal grip force.
Does ISO 10218:2025 guarantee a safe application?
No standard title alone guarantees safety. Apply the relevant requirements, assess application-specific hazards using ISO 12100 and other applicable guidance, implement risk reduction and verify the result.
Can a cobot project automatically receive BOI incentives?
No. The current measure has eligibility and investment conditions. Confirm the latest requirements with BOI and tax advisers before ordering equipment.
How many robot system integrators should be compared?
There is no fixed correct number. What matters is that candidates receive the same RFP, representative parts and FAT/SAT criteria. Compare relevant application experience, Thailand support, safety deliverables, change capability and every exclusion in the quotation.
Is a failed pilot a wasted investment?
No, provided success and stop conditions were defined in advance. Finding that feeding, tooling, safety, cycle time or demand makes the application unviable before a production-scale purchase is a valuable pilot result. Record the constraint and decide whether to redesign or stop.
Summary: one gate for safety, cycle time and TCO
Successful collaborative robot implementation is not defined by a robot brand or a fast demonstration. It is built by measuring the process, testing real parts, reducing lifecycle risk, comparing complete TCO, assigning responsibility in the RFP, and proving acceptance through FAT, SAT and a gated pilot. That is how a moving prototype becomes a maintainable production asset.
TOMAS TECH can support the early stage—from application screening and RFP structure to a practical safety, cycle-time and TCO review. You do not need a final robot model before starting the discussion. Share your current process and decision stage through our contact page.
Primary references
- ISO 10218-1:2025 — Industrial robots
- ISO 10218-2:2025 — Industrial robot applications and robot cells
- ISO/TS 15066:2016 — Collaborative robots
- Universal Robots — Risk assessment
- Universal Robots — ISO 10218 gets a makeover
- International Federation of Robotics — World Robotics 2025
- International Federation of Robotics — Robot density
- Thailand BOI — Smart and Sustainable Industry
- Thailand BOI — Q1 2026 investment applications
Note: This article reflects primary sources checked on 28 August 2026 and provides general procurement guidance, not legal, safety-certification, tax or investment advice. Confirm project-specific standards, laws, BOI conditions and safety with qualified professionals.
