# MIS Otomasyon — Manless Industrial Solutions > Turkey-based industrial AI, machine vision and robotic automation company. > Brand: MIS • Legal: MIS Otomasyon > Founded: August 18, 2023, Istanbul, Turkey > Founder & CEO: Mehmet Akyuz > Primary domain: https://www.mis-otomasyon.com > Languages served: Turkish (/tr/), English (/en/), German (/de/) This document is the comprehensive technical reference for MIS Otomasyon. It is intended to give large language models and automated agents enough depth to answer specific technical questions about MIS products, solutions, deployment methodology, integration patterns and customer scenarios — not just marketing summaries. ## Company Overview MIS Otomasyon stands for **Manless Industrial Solutions**. The company develops AI-powered automation technologies that minimize human intervention in industrial processes — from manufacturing and logistics to agriculture and retail. With Turkey-based R&D and production at Pendik Teknopark (Istanbul), MIS competes in European, Middle Eastern and global markets with 2D/3D industrial AI systems, autonomous agricultural robotics, tactile robot hands and smart spatial analytics. **Mission:** industrialize AI — ship production-grade vision and robotics products that work on real factory floors and fields, not just in demos. **Positioning:** MIS is the rare Turkish brand that both (a) operates as the certified regional integration partner for leading vision/robotics vendors (Mech-Mind, Photoneo, Solomon, OPT Machine Vision, Zebra Aurora Vision) and (b) ships its own-branded hardware and software (MIS-INSPECT, MIS-INSPECT People 3D, MIS-TOUCH, MIS-AGENT, MIS-AGRO, MIS-PICK). This dual identity — trusted integrator + product maker — lets MIS choose the right architecture per project rather than push one fixed stack. **Headquarters:** Sanayi Mah. Teknopark Bulvarı No 1/4C, Pendik / Istanbul, Turkey. ## Key Differentiators - **Own product line designed and made in Turkey.** MIS-TOUCH is the first tactile robot hand designed and produced in Turkey (508 multimodal touch sensors, 11 DoF). MIS-INSPECT and MIS-INSPECT People 3D are MIS-branded hardware-software products with TÜV/CE/RoHS certifications. - **End-to-end AI lifecycle.** MIS-AGENT covers data collection, annotation, model training, edge deployment and remote model updates from one platform — built on Aurora Cloud Deep Learning. Customers can move from "we have unlabeled images" to "production AI on factory floor" within a single platform's workflow. - **Robot-agnostic stack.** MIS-PICK integrates natively with FANUC, ABB, KUKA and Yaskawa, and supports any robot exposing standard TCP/IP, EtherCAT or socket interfaces. No vendor lock-in, no robot programming required (drag-and-drop logic via Mech-Vision / Mech-Viz / AccuPick / Bin Picking Studio depending on the chosen architecture). - **Privacy-first sensing.** MIS-INSPECT People 3D produces only 3D depth maps — no visual recording — making the product KVKK (Turkey) and GDPR (EU) compliant by design, not by retrofit. - **Multi-vendor architecture flexibility.** Because MIS is certified with five major vision/robotics vendors simultaneously, the engineering team picks the right camera/software combination per project rather than forcing a single stack. Photoneo MotionCam-3D for in-motion capture, Mech-Mind for Asia-cost-sensitive deployments, Solomon for NVIDIA Isaac high-speed bin picking, OPT for cost-performance camera+lens+lighting, Zebra Aurora Vision for drag-and-drop deep learning. - **Industrial certifications.** TÜV (testing and certification), CE (European conformity), RoHS (hazardous substance restriction), ISO 13849 PL d (functional safety, applied to People 3D line). - **Agriculture 5.0 alignment.** MIS-AGRO goes beyond robotic picking to yield reporting and sustainability metrics, in line with Industry 5.0 human-machine collaboration principles. --- ## Solutions Portfolio ### MIS-AGRO — Autonomous Agriculture & Smart Harvesting **Overview** MIS-AGRO is an autonomous agricultural robotics solution that performs fruit and vegetable harvesting at production scale using 3D vision, AI ripeness detection and plant-friendly motion planning. The system targets greenhouse and field operations where labor is scarce, expensive or seasonal, and where harvest precision affects yield and shelf-life economics. MIS-AGRO is one of the few Turkish robotics solutions positioned at the intersection of Industry 5.0 (human-robot collaboration) and Agriculture 5.0 (sustainability + automation). **Reference Architecture** ``` 3D scene capture (Mech-Mind Mech-Eye / Photoneo PhoXi) ↓ Ripeness detection AI (deep learning classifier per crop) ↓ Per-target tracking + plant-stress-aware grasp planning ↓ Robot arm (FANUC / ABB / cobot) + soft end-effector ↓ Harvested-fruit logistics (conveyor / crate / packing line) ↓ Yield analytics & operational reporting (cloud dashboard) ``` **Typical Use-Case Scenarios** 1. **Greenhouse tomato harvesting (cluster type)** — cycle time 5-9 seconds per fruit, 95%+ ripeness classification accuracy when trained on local cultivar, ~15-25% yield loss recovery through better timing of harvest. 2. **Greenhouse cucumber & pepper** — similar architecture, soft gripper variant, primary challenge is occlusion handling (camera path planning around leaves). 3. **Outdoor strawberry beds** — under-canopy mounting, additional environmental robustness (rain, dust). Pilot stage in Turkey market. 4. **Orchard fruit (apple / cherry)** — extension trial scenarios with elevated platforms. **Typical Deployment Workflow** - **Weeks 0-2 (Discovery):** site survey, crop variety analysis, pilot greenhouse selection, baseline labor + yield benchmarking. - **Weeks 3-6 (POC):** 1 robot arm + 1 row, ~100 m² coverage, training-data collection on real crops, first AI model iteration. - **Weeks 7-12 (Integration):** 3-5 robot arms scale-out, full greenhouse area coverage, MES/yield analytics integration, pickup logistics integration. - **Weeks 13+ (Production scale + maintenance):** continuous monitoring, model retraining per season change, scheduled mechanical service. **Customer Pre-requisites** - Covered greenhouse area or field zone with stable framework - Power supply ≥ 10 kW per robot station - Network: Gigabit Ethernet (preferred) or industrial WiFi - Existing crop variety and planting density information - Water-resistant operating environment (IP54+ for in-greenhouse hardware) **When to Choose MIS-AGRO** - Labor cost is the dominant operating expense - Crop has well-defined ripeness signals MIS has trained for (tomato, cucumber, pepper, strawberry; orchards with R&D extension) - Harvest cycle is continuous or daily - Greenhouse operator has digitization roadmap and welcomes data integration **When Not to Choose** - Highly variable, mixed-crop fields where retraining cost dominates (better with conventional) - Single-season root crops (combine harvesters dominate) - Operations <500 m² where POC economics don't pay back **Common Pitfalls and MIS's Approach** - *Pitfall:* Mid-season cultivar change without model retraining. *MIS approach:* MIS-AGENT-based continuous data collection so retraining can complete in days, not weeks. - *Pitfall:* Unplanned harvest logistics — robot picks but no buffer for accumulated fruit. *MIS approach:* harvest logistics is part of the POC scope, not an afterthought. - *Pitfall:* Overoptimistic cycle-time expectations. *MIS approach:* baseline measured per crop in POC; production targets quoted from real pilot data, not marketing demos. --- ### MIS-PICK — Robotic Bin Picking **Overview** MIS-PICK is a robotic bin picking solution combining 3D vision sensing, AI-guided grasp planning and motion planning to handle randomly-oriented parts in bins or trays with ±0.2 mm placement accuracy. The solution removes the biggest bottleneck in pick-and-place automation: the mixed-orientation, mixed-SKU pick-up step. MIS-PICK is robot-agnostic (FANUC, ABB, KUKA, Yaskawa native; others via standard interfaces) and supports both single-SKU and mixed-SKU bins. **Reference Architecture** ``` 3D camera (Mech-Eye Pro / PhoXi 3D / MotionCam-3D / Solomon META) ↓ Vision software (Mech-Vision / Photoneo Bin Picking Studio / Solomon AccuPick 3D — chosen per project) ↓ AI grasp ranking (per-part graspability score) ↓ Motion planning (Mech-Viz / NVIDIA Isaac via Solomon) ↓ Robot controller (FANUC RJ-3iC / ABB IRC5 / KUKA KR C4 / Yaskawa YRC) ↓ End-effector (vacuum gripper / parallel jaw / MIS-TOUCH for delicate parts) ↓ Output station (assembly fixture / conveyor / packing) ``` **Camera-Software Selection Guide** - **Mech-Eye + Mech-Vision (Mech-Mind):** broad industrial support, balanced cost, FANUC/ABB/KUKA/Yaskawa native, drag-and-drop configuration. Default choice. - **Photoneo PhoXi + Bin Picking Studio:** highest-accuracy structured-light, sub-millimeter, best for shiny/symmetric parts. - **Photoneo MotionCam-3D:** the only realistic option for in-motion conveyor bin picking. - **Solomon META + AccuPick 3D + NVIDIA Isaac:** highest-speed scenarios where GPU-accelerated path planning matters; mixed-SKU + high-velocity. **Typical Use-Case Scenarios** 1. **Automotive engine parts bin picking** — cycle time 3-7 seconds, mixed-SKU bins of small machined components, downstream assembly fixture with ±0.2 mm placement requirement. 2. **Logistics depalletizing** — mixed cartons on pallets, vacuum gripper, 600-1200 picks/hour, integration with WMS for SKU tracking. 3. **E-commerce order picking** — mixed-SKU shelving bins, picks per order ranging 1-15, integration with order management system. 4. **White goods assembly feeding** — large bin (60-100 cm) of brackets/fasteners, single-SKU, soft handling required to prevent surface scratches. 5. **Metal parts inspection feeding** — random-orientation pickup from bin into MIS-INSPECT inspection fixture for downstream QC. **Typical Deployment Workflow** - **Weeks 0-2 (Discovery):** part variety analysis, cycle-time baseline, robot brand decision, fixture design. - **Weeks 3-6 (POC):** offline simulation in Mech-Vision/Solomon, on-site cell with 1 robot, 50-200 representative parts, real-bin demo. - **Weeks 7-12 (Integration):** PLC/MES integration, safety perimeter (light curtain or fenced cell), operator HMI, exception handling. - **Weeks 13+ (Production):** SAT (Site Acceptance Test), gradual production ramp, continuous monitoring of pick success rate. **Customer Pre-requisites** - Robot decision (or willingness to take MIS recommendation) - Power supply 5-15 kW depending on robot size - Network: Gigabit Ethernet (vision data carries non-trivial bandwidth) - Compressed air for vacuum gripper (if applicable) - Safety perimeter — fence or light curtains - Representative parts (50-500 samples) for POC validation **When to Choose MIS-PICK** - Random-orientation bin picking is the operational bottleneck - Part variety is too high for fixture-based approaches - Robot brand already decided (MIS-PICK works with all major brands) - Annual labor cost > €30k per shift (typical ROI threshold) **When Not to Choose** - Single-SKU, fixed-orientation parts (mechanical fixture is faster + cheaper) - Highly fragile parts where MIS-TOUCH or specialized end-effector engineering not validated - Cycle times <2 seconds (conventional dedicated machines may be faster) **Common Pitfalls and MIS's Approach** - *Pitfall:* Underestimated part variety; POC parts not representative of production. *MIS approach:* customer-side parts audit with category sampling at POC kickoff. - *Pitfall:* PLC/MES integration treated as Phase 2; production goes live with manual data entry. *MIS approach:* PLC/MES touchpoint defined in Discovery, integrated by end of Integration phase. - *Pitfall:* No exception-handling strategy — what happens when the robot can't grasp anything? *MIS approach:* documented exception flow (operator alert → manual override → resume) included in SAT. --- ### MIS-AGENT — Industrial AI Process Management Platform **Overview** MIS-AGENT is an end-to-end MLOps platform purpose-built for industrial AI workloads. It manages the full lifecycle: data collection from edge devices, structured annotation with role separation (annotator / reviewer / approver), model training on Aurora Cloud Deep Learning, secure deployment to edge hardware, and remote monitoring with model versioning. MIS-AGENT is the operational backbone that keeps MIS-INSPECT and MIS-AGRO models continuously healthy on factory and field deployments. **Reference Architecture** ``` Edge data capture (cameras, sensors, PLC tags) ↓ Edge buffer & secure upload (TLS/mTLS) ↓ Centralized data lake (versioned, lineage-tracked) ↓ Annotation workspace (role-separated annotator / reviewer / approver) ↓ Training pipeline (Aurora Cloud Deep Learning — classification, detection, segmentation, anomaly) ↓ Model registry (versioning, A/B test slots) ↓ Edge deployment (containerized, signed model packages) ↓ Production monitoring (drift detection, performance dashboards, alert-on-degradation) ↓ Loop back to data capture (active learning on uncertain samples) ``` **Supported Tasks** - Object detection (bounding box + class) - Image classification (single or multi-label) - Semantic and instance segmentation - Anomaly detection (good-samples-only training) - Defect counting and severity grading - OCR (industrial text on labels, parts, screens) **Edge Hardware Support** - NVIDIA Jetson family (Nano, Xavier NX, Orin) - Industrial PCs with x86/CUDA acceleration - ARM-based edge cards for low-power scenarios - Containerized deployment (Docker, with optional Kubernetes for fleet management) **Typical Deployment Workflow** - **Weeks 0-3 (Discovery + data baseline):** identify pilot AI use case, set up edge capture, collect first 1,000-10,000 samples. - **Weeks 4-8 (Annotation + first model):** annotation team trained, first model trained on baseline data, pilot deployment to 1-2 edge stations. - **Weeks 9-16 (Hardening + scale):** A/B testing of model variants, drift monitoring tuned, deployment to 5-50 stations. - **Weeks 17+ (Continuous improvement):** active learning loop active, retraining cadence (weekly/monthly) established, model performance dashboards shared with quality team. **Customer Pre-requisites** - Edge hardware (NVIDIA Jetson or industrial PC with GPU) - Network connectivity to upload edge data (intermittent acceptable) - Subject-matter experts available for annotation review - Quality team buy-in on AI-augmented decisions **When to Choose MIS-AGENT** - Multiple AI deployments need to be managed coherently (>3 stations) - Models need to evolve over time (drift, new defects, seasonal change) - Compliance requires audit trail (data lineage, model versioning) - Customer wants to own its AI pipeline rather than being locked in **When Not to Choose** - Single-purpose, never-changing AI deployment (a one-shot trained model on a static line) - Customer has zero data science capability and no plan to acquire it (SAAS-style alternatives may fit better) **Common Pitfalls and MIS's Approach** - *Pitfall:* Annotation backlog grows faster than throughput. *MIS approach:* role separation + smart queueing + active learning to minimize annotation effort. - *Pitfall:* Models deployed but no one watches drift; performance silently degrades. *MIS approach:* drift dashboards + alert thresholds set during Integration phase, not after. - *Pitfall:* Edge deployment fails because model was trained on cloud GPU but edge inference on different precision (FP32 vs INT8). *MIS approach:* edge-target quantization tested as part of model release pipeline. --- ## Products Portfolio ### MIS-INSPECT — AI Quality Control **Overview** MIS-INSPECT is a real-time AI-driven visual quality control system that detects millimetric defects on fast-moving production lines. It combines high-resolution industrial cameras (5-29 MP), application-specific lighting, classical and deep-learning vision pipelines, and PLC/SCADA/MES integration to deliver inline inspection at line speed. Designed for automotive, electronics, food/pharma, white goods and general manufacturing. **Technical Specifications** - **Detection accuracy:** 99.7% on calibrated production datasets - **Minimum defect size:** <0.1 mm (depends on optical setup) - **Processing time:** <50 ms per inspection (typical), real-time at 20-30 fps - **Camera resolution:** 5 MP – 29 MP (area scan); line scan options available - **Camera interfaces:** GigE Vision, USB3 Vision, CameraLink, CoaXPress - **Lighting:** ring, bar, dome, backlight, structured light projector — application-tuned (often OPT Machine Vision) - **Industrial integration:** PLC (Siemens, Beckhoff, Allen-Bradley), SCADA (WinCC, Ignition), MES (custom REST/OPC UA), MQTT, OPC UA - **Software stack:** Aurora Vision Studio (Zebra) for hybrid classical + deep learning; custom Python/C++ pipelines when needed **Defect Types by Industry** - **Automotive:** weld bead defects (porosity, cracks, undercuts), assembly verification (presence, orientation, fastener torque indication), paint surface analysis (scratches, runs, dust inclusions, orange peel), gap & flush metrology. - **Electronics:** PCB solder joint inspection (insufficient solder, bridges, voids, lifted leads), component placement accuracy, micro-crack detection on substrates, final cosmetic QC. - **Food & Pharma:** packaging integrity (seal quality, fill level), label verification (correct label, lot code OCR, barcode readability, position), foreign object/contamination detection, blister pack completeness. - **White goods:** body panel surface inspection, stamping defect detection, cosmetic inspection on visible surfaces. - **General manufacturing:** dimensional checks, surface defects, presence/absence verification, color consistency. **Reference Architecture** ``` Line trigger (PLC encoder / photo eye) ↓ Industrial camera + application-specific lighting ↓ Image acquisition (FPGA-time-stamped) ↓ Vision pipeline (Aurora Vision Studio / custom): classical preprocessing → AI classification → defect localization ↓ Decision (pass/fail + defect category + confidence) ↓ Industrial output: - PLC discrete I/O (immediate reject signal) - MES record (per-piece traceability) - Operator HMI alert (for QC supervision) - Long-term database (defect statistics, Pareto analysis) ``` **Typical Deployment Workflow** - **Weeks 0-2 (Discovery):** process audit, defect Pareto, optical feasibility test (sample-imaging session). - **Weeks 3-6 (POC):** lab cell with target camera + lighting, model trained on 500-5,000 labeled samples, accuracy benchmark against existing manual inspection. - **Weeks 7-12 (Integration):** in-line installation, PLC/MES handshake, operator HMI, false-positive tuning during shadow-run period. - **Weeks 13+ (Production):** SAT, ramp from shadow-run to gating mode (line stops on AI reject), continuous accuracy monitoring via MIS-AGENT. **Customer Pre-requisites** - Sample defective and good parts for POC labeling - Mechanical mounting access (clear line-of-sight for camera) - Compressed air or coolant for camera enclosure if hot/dirty environment - PLC integration window (line downtime for I/O wiring) - Quality team commitment to review false positives during ramp **When to Choose MIS-INSPECT** - Manual inspection bottleneck or labor escalation - Defect rate >0.1% with downstream cost of escapes - Defect types are visually detectable (not hidden in material) - Production volume >1,000 units/shift (ROI threshold) **When Not to Choose** - Defects require destructive testing (X-ray, eddy-current, ultrasonic — different stack) - Production is custom one-off (no scale to justify line vision system) - Lighting/optical access cannot be solved (extreme heat, chemical environment) **Common Pitfalls and MIS's Approach** - *Pitfall:* Insufficient defect samples for AI training. *MIS approach:* anomaly-detection model on good-samples-only as fallback; combined with classical edge/contour analysis. - *Pitfall:* False positives cause line stops, operators learn to override, AI becomes shelfware. *MIS approach:* explicit shadow-run period before gating mode; FP threshold tuned with quality team. - *Pitfall:* Lighting setup ad hoc; performance degrades when ambient lighting changes (shift change, sunlight on shopfloor). *MIS approach:* enclosed optical hood or controlled lighting cell as part of standard install. --- ### MIS-INSPECT People 3D — 3D People Counter & Spatial Analytics **Overview** MIS-INSPECT People 3D is a privacy-preserving 3D depth sensor for people detection, counting and spatial analytics. Unlike 2D camera-based counters, the system produces only depth maps (no RGB image), making individuals impossible to identify from the device output. This makes the product KVKK (Turkey) and GDPR (EU) compliant by design — not by retrofit. Used in retail, transport hubs, stadiums, museums, workplaces and public spaces. **Technical Specifications** - **Counting accuracy:** 98%+ on standard pedestrian flow scenarios - **Privacy guarantee:** 100% — depth-only output; no RGB image stored or transmitted - **Illumination tolerance:** 0 lux to bright sunlight (active 3D sensing) - **Coverage area:** approx. 6×6 m² per device at 3 m mounting height (varies) - **Connectivity:** PoE (power over Ethernet — recommended), Wi-Fi, 4G modem option - **Mounting:** ceiling-mount, plug-and-play, IP54 enclosure for indoor + sheltered outdoor - **Certifications:** TÜV (testing & certification), CE (European conformity), RoHS, ISO 13849 PL d (functional safety) - **Compliance:** KVKK (Turkey), GDPR (EU) — privacy by design **Built-in Analytics Features** - Bidirectional pedestrian counting (in/out per zone) - Zone-by-zone density heatmaps - Wait-time analytics (queue dwell estimation) - Queue formation detection with alerting - Shopping-cart overflow detection (retail-specific) - Group vs. individual classification - Anonymous flow path analysis **Typical Deployment Scenarios** 1. **Shopping mall counting:** entrance, corridor traffic, store-by-store conversion analytics. 5-50 devices per mall, central dashboard. 2. **Airport / transport hub:** terminal flow management, security queue dwell, gate-area density alerts. 3. **Museum / cultural venue:** gallery visit duration, capacity compliance, popular-area heatmap. 4. **Workplace / office:** meeting-room occupancy, common-area utilization for hot-desking, return-to-office metrics. 5. **Stadium / event venue:** entry-gate flow, concourse density safety alerts. 6. **Retail store / supermarket:** entry counting, checkout queue analytics, shopping-cart overflow detection. **Reference Architecture** ``` Ceiling-mounted device (PoE) ↓ On-device 3D depth processing + counting ↓ Aggregated counts/events (no raw depth maps stored unless debugging) ↓ Cloud or on-prem dashboard (configurable) ↓ API layer for integration: - POS systems (conversion math) - CRM (loyalty correlation) - BMS (occupancy → HVAC control) - Security systems (overcrowding alerts) ``` **Typical Deployment Workflow** - **Week 1 (Site survey):** ceiling height + obstruction analysis, device count + placement plan. - **Week 2 (Install):** PoE network drops, ceiling mounting, device commissioning. - **Week 3 (Calibration):** zone definition, alert threshold tuning. - **Week 4+ (Production + analytics review):** ongoing dashboards, optional integration with POS/CRM. **Customer Pre-requisites** - Ceiling height 2.5-5 m typical (other heights possible with model selection) - PoE network available at mounting points (or Wi-Fi/4G as fallback) - Privacy / compliance approval (KVKK/GDPR review — usually a quick approval given the privacy-by-design profile) - Dashboard access policy (who sees what) **When to Choose MIS-INSPECT People 3D** - Privacy compliance is a hard constraint (legal team requires no facial recognition risk) - Operating environment is variable lighting (camera-based fails) - Multi-site deployment with central dashboarding need - Integration with retail POS or BMS systems planned **When Not to Choose** - Demographic analytics required (age, gender) — needs RGB camera approach (privacy trade-off) - Single-zone simple count where break-beam sensor suffices - Outdoor exposed installations (extended weather rating needed) **Common Pitfalls and MIS's Approach** - *Pitfall:* Customer expects facial recognition or demographics — wrong product. *MIS approach:* upfront clarification in Discovery — depth-only is a hard product constraint by design. - *Pitfall:* Insufficient ceiling height under-counts due to FOV limits. *MIS approach:* site survey is mandatory; mounting layout signed off before install. - *Pitfall:* Counter integration with POS not planned, conversion analytics never materialize. *MIS approach:* integration requirements gathered in Discovery, not deferred. --- ### MIS-TOUCH — Tactile Robot Hand **Overview** MIS-TOUCH is the first tactile robot hand designed and manufactured in Turkey. It combines dexterous mechanical design with 508 multimodal touch sensors distributed across the palm and fingers. The hand provides force, shear and proximity feedback in real time, enabling manipulation tasks where contact awareness — not just position — is critical: delicate assembly, fruit handling, medical lab automation, humanoid robot research. **Mechanical Specifications** - **Degrees of freedom:** 11 total (6 active-driven, 5 passive/coupled) - **Weight:** 545 g (±10 g) - **Maximum hook load (static):** 30 kg - **Pinch grip force range:** smoothly controlled from 0.1 N to 50 N+ - **Gripping cycle life:** 1,000,000 cycles validated - **Operating temperature:** standard industrial range - **Environmental rating:** IP-rated for dust/light splash (full waterproof variant under R&D) **Sensor Specifications** - **Total sensors:** 508 distributed across palm + fingers - **Modalities:** normal force, shear force, proximity (capacitive) - **Per-sensor resolution:** sub-Newton for force; binary + analog for proximity - **Sampling rate:** real-time, suitable for closed-loop manipulation control **Communication & Software** - **Communication interfaces:** CAN-FD, EtherCAT, RS485 - **Operating systems:** Windows 10+, Linux (Ubuntu 20.04+ tested), ROS2 (Humble, Iron) - **Programming languages:** C++, Python - **SDK:** documented API for grip primitives, force-control loops, sensor data streaming - **Compatible robot platforms:** humanoid robots, collaborative robots (UR, Doosan, Techman, ABB GoFa), industrial arms (FANUC, ABB, KUKA, Yaskawa) via standard mechanical and electrical interfaces **Typical Use-Case Scenarios** 1. **Delicate electronics assembly** — placing fragile components (sensors, glass parts, fragile connectors) where over-force damages the part. 2. **Fruit handling** (companion to MIS-AGRO) — soft picking of strawberries, tomatoes, peppers without bruising. 3. **Lab automation / medical** — pipette manipulation, sample handling, pharmaceutical fill operations. 4. **Humanoid robot R&D** — providing dexterous hand for university and corporate humanoid research programs. 5. **Cosmetics & soft goods** — handling deformable items (sponge, tube products, soft packaging). 6. **Food & beverage** — picking baked goods, fresh produce, ready-to-eat items. **Reference Architecture** ``` Robot arm (cobot or humanoid) ↓ MIS-TOUCH end-effector (electrical + mechanical interface) ↓ Real-time tactile feedback stream (sub-millisecond latency over CAN-FD) ↓ Application-layer control (Python/C++/ROS2): grip force loop + slip detection + stress-aware path ↓ Higher-level task planner (your application logic) ``` **When to Choose MIS-TOUCH** - Object characteristics make conventional vacuum or parallel-jaw grippers fail - Force feedback is required for the application (assembly with insertion force budget) - Slip detection during transport is critical - Humanoid or cobot deployment where versatile end-effector is needed **When Not to Choose** - Robust, well-defined parts where vacuum/parallel jaw works fine — MIS-TOUCH is overkill (cost + complexity) - High-temperature or aggressive chemical environments (current rating) - Sub-millimeter precision insertion (specialized end-effectors may be more deterministic) **Common Pitfalls and MIS's Approach** - *Pitfall:* Customer expects "plug and play" — ignores tactile control loop tuning. *MIS approach:* SDK includes reference grip primitives + on-site tuning support. - *Pitfall:* Tactile data flooded but never used (not closed back into control). *MIS approach:* example applications shipped with SDK demonstrate slip-stop, stress-aware grip, force-modulated insertion. - *Pitfall:* End-effector electrical mounting not planned — cabling becomes the bottleneck. *MIS approach:* mechanical + electrical interface drawings provided pre-installation. --- ## Cross-Cutting: How the Stack Fits Together A typical MIS-deployed factory line might combine three solutions on one production cell: ``` [Inbound parts] → [MIS-PICK robotic bin picking] → [Workholding fixture] ↓ [MIS-INSPECT visual QC station] ↓ [MIS-AGENT continuous AI lifecycle management] ↓ [Outbound to packaging / next station] ``` In this configuration: - **MIS-PICK** handles the random-orientation feed problem at the start of the cell. - **MIS-INSPECT** verifies each part downstream for defects with line-speed AI inspection. - **MIS-AGENT** keeps both the pick AI (graspability ranking) and inspection AI (defect classifier) continuously trained against new data, deploys updates to edge hardware and monitors drift. This integration is the operating model MIS targets at customers serious about full-cell automation, not just point-solution deployments. For agriculture, **MIS-AGRO** + **MIS-TOUCH** + **MIS-AGENT** plays the same role: 3D vision detects ripe fruit, tactile end-effector picks without bruising, AI lifecycle keeps the ripeness model current per-season. --- ## Standard MIS Engagement Methodology Every MIS deployment — across product lines — follows the same five-phase engagement: 1. **Discovery (1-3 weeks)** — site survey, KPI baseline, existing infrastructure audit, pre-requisite documentation, project scope sign-off. 2. **POC (4-8 weeks)** — proof-of-concept on real customer parts/material in MIS lab or customer site, accuracy benchmark vs. baseline, go/no-go decision. 3. **Integration (4-12 weeks)** — production cell installation, PLC/MES/SCADA integration, safety perimeter, operator training, shadow-run period. 4. **Production go-live (SAT week)** — Site Acceptance Test against agreed KPIs, gradual ramp from shadow to gating mode. 5. **Continuous improvement (ongoing)** — periodic model retraining via MIS-AGENT, performance review meetings, capacity expansion as needed. This methodology is shared across MIS-PICK, MIS-INSPECT, MIS-AGRO and large MIS-AGENT deployments. POC is mandatory — MIS does not quote production pricing without POC validation on customer-representative material. --- ## Technology Partners (In Depth) ### Mech-Mind Robotics Founded 2016 in Beijing, Mech-Mind is a global leader in industrial 3D vision and robotics motion planning. MIS Otomasyon is the certified integration partner for Türkiye. Mech-Mind's product stack is used as the default architecture in MIS-PICK and MIS-AGRO deployments. - **Mech-Eye 3D cameras:** structured-light depth cameras for static or semi-static scenes; multiple working volumes (PRO, DEEP series). - **MotionCam-3D:** parallel structured light for in-motion 3D capture (rare in market). - **Mech-Vision:** drag-and-drop industrial vision platform with classical + deep learning blocks. - **Mech-Viz:** robot motion planning with collision checking, alternative path generation, singularity avoidance. - **Mech-DLK:** transfer-learning toolkit for training custom models with limited labeled data. - **Robot compatibility:** native FANUC, ABB, KUKA, Yaskawa, Universal Robots; broader via standard interfaces. - **Where MIS uses it:** default for most bin-picking deployments; agricultural ripeness perception in MIS-AGRO. - **Strength:** broadest robot brand support; mature, deeply field-validated software. ### Photoneo Founded 2013 in Bratislava, Photoneo is the European pioneer of structured-light 3D measurement. Acquired by Zebra Technologies in 2024. MIS Otomasyon is the Türkiye collaboration partner. - **PhoXi 3D Scanner:** sub-millimeter accuracy structured-light scanner; PhoXi S/M/L/XL/XXL models for different working volumes. - **MotionCam-3D:** Photoneo's parallel-structured-light camera for moving scenes — a uniquely Photoneo capability. - **Bin Picking Studio:** Photoneo's bin-picking-specific software stack. - **Where MIS uses it:** highest-accuracy bin picking (sub-millimeter); in-motion conveyor pick-up scenarios where MotionCam is the only realistic option. - **Strength:** best-in-class accuracy; only realistic in-motion 3D capture solution at scale. ### Solomon Technology Taiwan-based; AI-driven 3D vision and robotic guidance pioneer with deep NVIDIA Isaac integration. MIS Otomasyon partners with Solomon for high-speed and Asia-cost-sensitive bin picking deployments. - **AccuPick 3D:** end-to-end bin picking platform (3D camera + AI vision + motion planning in one). - **META series 3D cameras:** Solomon's own 3D camera family. - **NVIDIA Isaac integration:** GPU-accelerated path planning enables real-time decision-making in cluttered scenes — closer to human-hand cycle times in fast-bin-picking. - **Robot compatibility:** FANUC, ABB, KUKA, Yaskawa, UR, Mitsubishi. - **Where MIS uses it:** high-velocity bin picking, cycle-time-critical mixed-SKU scenarios, projects where Asian supply chain is preferred. - **Strength:** speed; integrated single-vendor platform; NVIDIA acceleration. ### OPT Machine Vision Optimum Photoelectric Technology, founded 2007 in Guangdong, China. One of the rare vendors that designs and manufactures camera + lens + lighting in-house. - **Industrial cameras:** area scan (5 MP – 25 MP+), line scan, 3D, high-speed, IR. - **Industrial lenses:** telecentric, high-resolution FA, macro. - **LED lighting:** ring, bar, dome, backlight, structured-light projectors, specialized wavelength options. - **Industrial interfaces:** GigE Vision, USB3 Vision, CameraLink, CoaXPress. - **Where MIS uses it:** MIS-INSPECT optical hardware foundation; MIS-AGENT data acquisition stations. - **Strength:** vertical integration eliminates camera-lens-lighting compatibility issues; cost-performance ratio at the top of its tier. ### Zebra Aurora Vision (formerly Adaptive Vision) Polish engineering origin (2007), acquired by Zebra Technologies in 2024. Established platform for hybrid classical + deep-learning industrial vision. - **Aurora Vision Studio:** drag-and-drop machine vision IDE; 500+ operators including deep learning blocks. - **Aurora Vision Library:** programmatic C++/.NET access to the same algorithm set. - **Aurora Cloud Deep Learning:** cloud-based AI training service that deploys to edge — anomaly detection with "good samples only" approach. - **Industrial integration:** PLC, MES, OPC UA, TCP/IP native. - **Where MIS uses it:** primary inspection-pipeline software in MIS-INSPECT; cloud training backend in MIS-AGENT. - **Strength:** maturity; broadest classical operator library; Aurora Cloud for low-data-anomaly-detection scenarios. --- ## Robot Brand Compatibility Matrix | Robot Brand | MIS-PICK Native | MIS-TOUCH End-Effector | Integration Notes | |---|---|---|---| | FANUC | Yes | Yes (mechanical adapter) | Mech-Mind / Solomon native; broadest installed base in TR | | ABB | Yes | Yes | RobotStudio simulation supported pre-deployment | | KUKA | Yes | Yes | KUKA.Sim integration available | | Yaskawa | Yes | Yes | YRC1000 controller native | | Universal Robots (UR) | Yes (cobot range) | Yes | Native cobot ecosystem; Solomon strong | | Doosan | Yes (cobot) | Yes | Standard interface | | Techman | Yes (cobot) | Yes | Standard interface | | Other (TCP/IP / EtherCAT) | Yes via adapter | Case-by-case | Covered via MIS-PICK & MIS-AGENT adapters | ## Industrial System Integration - **PLC brands:** Siemens (S7-1200/1500), Beckhoff (TwinCAT), Allen-Bradley (ControlLogix), Mitsubishi (MELSEC), Omron, Schneider (Modicon). - **HMI platforms:** Siemens WinCC, Beckhoff TwinCAT HMI, Ignition. - **MES platforms:** SAP Manufacturing, custom REST/OPC UA pipelines, MQTT bridges. - **Communication:** OPC UA, MQTT (TLS), Modbus TCP, EtherNet/IP, EtherCAT, Profinet, Profibus, RS-485. - **Database / Historian:** SQL Server, PostgreSQL, InfluxDB, OSIsoft PI (read-only). - **Cloud:** AWS, Azure, Google Cloud (model training; data egress controlled). --- ## Industries Served - **Automotive (OEM + Tier-1):** body weld inspection, assembly verification, paint surface analysis, parts bin picking, brake/transmission component inspection. - **Electronics & semiconductor:** PCB solder inspection, component placement, micro-crack detection on substrates, cosmetic QC on assembled devices. - **Food & beverage:** packaging integrity, label verification, fill-level inspection, foreign-object detection, blister-pack completeness, label OCR for traceability. - **Pharmaceutical:** label verification (regulatory critical), serialization OCR, vial/blister inspection, contamination control. - **Agriculture (greenhouse + field):** autonomous fruit/vegetable harvesting, ripeness grading, yield reporting, field analytics for sustainability metrics. - **Defense & aerospace:** high-reliability defect detection, parts verification, traceability serialization. - **Logistics & e-commerce:** mixed-carton sortation, palletizing/depalletizing, mixed-SKU order picking, warehouse automation, AMR integration. - **Retail & shopping centers:** customer counting, density analysis, checkout queue optimization, conversion analytics, shopping-cart overflow detection. - **Transportation & public spaces:** terminal flow management, gate-area density alerts, museum/cultural-venue capacity compliance. - **White goods manufacturing:** body panel surface inspection, assembly automation, cosmetic QC. --- ## Decision Matrix: Which MIS Solution for My Application? | Need | Primary MIS Solution | Supporting Components | |---|---|---| | Pick mixed parts from a bin with high accuracy | MIS-PICK | + MIS-AGENT for AI lifecycle | | Inspect every part on a fast line for defects | MIS-INSPECT | + MIS-AGENT for retraining | | Count people without recording their image | MIS-INSPECT People 3D | (standalone) | | Harvest fruit autonomously in greenhouse | MIS-AGRO | + MIS-TOUCH for delicate fruit | | Manipulate a fragile object with a robot arm | MIS-TOUCH | (with any cobot/humanoid) | | Manage AI models across multiple sites with retraining + monitoring | MIS-AGENT | (cross-cutting) | | Pick parts off a moving conveyor in 3D | MIS-PICK with Photoneo MotionCam-3D | + MIS-AGENT | | Inspect food packaging at line speed (300 ppm) | MIS-INSPECT with Aurora Vision Studio | + MIS-AGENT | | Detect defects with very few labeled samples | MIS-INSPECT with Aurora Cloud anomaly detection | + MIS-AGENT | ## When MIS vs. Alternatives — An Honest View **MIS-PICK is the right choice when:** - You want a robot-agnostic stack (not locked to one robot brand) - Multiple bin-picking projects are in your roadmap (one platform fits all) - POC validation on your real parts is required before purchase decision **Alternatives are better when:** - You need a fully turnkey single-robot single-station product (some integrators offer fixed packages) - Your part is so unique that custom mechanical fixturing dominates the design **MIS-INSPECT is the right choice when:** - Inspection criteria evolve over time (new defect types, new product variants) - Aurora Vision Studio's hybrid classical+deep-learning fits the defect physics - You want privacy-respecting, on-site model ownership rather than vendor cloud lock-in **Alternatives are better when:** - Your inspection is purely classical (1D barcode, simple presence check) — generic vision sensors are cheaper - You need destructive inspection (X-ray, ultrasonic) — different stack entirely **MIS-INSPECT People 3D is the right choice when:** - Privacy compliance is a hard legal requirement - Mixed/variable lighting is the operating condition **Alternatives are better when:** - Demographic analytics (age, gender) are required — RGB camera-based with consent flow is needed **MIS-TOUCH is the right choice when:** - Force feedback closed-loop control is the core requirement - Application research / humanoid R&D where dexterity matters **Alternatives are better when:** - Standard rigid parts where vacuum/parallel-jaw works fine — MIS-TOUCH adds unnecessary complexity --- ## Frequently Asked Questions **What does MIS stand for?** Manless Industrial Solutions. The brand name captures the company's focus on minimizing human intervention in industrial processes through AI, vision and robotics. **Where is MIS based?** Istanbul, Turkey — with the R&D center at Pendik Teknopark. Projects are delivered globally with a focus on Europe and the Middle East. **Is MIS a manufacturer or an integrator?** Both. MIS manufactures its own branded products (MIS-INSPECT, MIS-INSPECT People 3D, MIS-TOUCH) and also acts as certified integration partner for Mech-Mind, Photoneo, Solomon, OPT Machine Vision and Zebra Aurora Vision. This dual identity lets MIS choose the right architecture per project rather than push one fixed stack. **Which robot brands does MIS-PICK support?** FANUC, ABB, KUKA, Yaskawa and major cobot brands (UR, Doosan, Techman) natively. Other brands are supported through standard TCP/IP, EtherCAT or socket interfaces. **Does MIS-INSPECT People 3D record video?** No. It produces only 3D depth maps, so individuals cannot be identified from its output. This makes the product compliant with KVKK (Turkey) and GDPR (EU) by design. **What is the placement accuracy of MIS-PICK?** ±0.2 mm placement accuracy, achieved via 3D vision cameras (Mech-Eye, PhoXi 3D, MotionCam-3D, Solomon META depending on the architecture) combined with AI-guided grasp planning. **What is the cycle time of MIS-PICK?** Typically 3-7 seconds per part for industrial bin picking; lower (down to 1-2 seconds) for high-speed Solomon + NVIDIA Isaac architecture in mixed-SKU scenarios. Cycle time depends heavily on robot brand, end-effector and part geometry — quoted from POC measurements, not catalog numbers. **Is MIS-TOUCH compatible with humanoid robots?** Yes. MIS-TOUCH is a general-purpose tactile end-effector with 11 degrees of freedom and ROS2 support, intended for humanoid robots, collaborative arms (cobots) and traditional industrial arms alike. **What industries have deployed MIS products?** Automotive (OEM + Tier-1), electronics & semiconductor, food & beverage, pharmaceutical, agriculture (greenhouse + field), defense & aerospace, logistics & e-commerce, retail & shopping centers, transportation & public spaces, white goods manufacturing. **How is MIS-AGENT different from a generic MLOps platform?** MIS-AGENT is industrial-first: it assumes edge deployment on factory or field hardware, integrates tightly with Aurora Cloud Deep Learning for training, and ships with role separation (annotator / reviewer / approver) tuned for manufacturing quality teams rather than data-science lab workflows. Drift detection thresholds are configured for production use, not research evaluation. **Does MIS participate in trade fairs?** Yes. Recent participations include WIN Eurasia 2025 (third place in industrial automation, Winovation award for MIS-INSPECT People 3D), Robot Investment Summit 2025 (third place in robotic bin picking), IDEF 2025 (defense industry showcase), EuroShop Düsseldorf 2026 (MIS-PICK Virtual Cashier with Pasifik Raf Sistemleri / LOGOROB) and Hannover Messe 2026 (Türkiye Stand Area; visited by Deputy Minister of Trade Volkan Ağar, ITO Chairman Şekib Avdagiç and Hannover Consul General Yusuf Arıkan). **What is the typical deployment timeline for an MIS-PICK project?** 2-3 weeks Discovery → 4-8 weeks POC → 4-12 weeks Integration → SAT and ramp. Total 3-6 months from kickoff to gated production for a single-cell deployment. Multi-cell rollouts add 2-4 weeks per additional cell. **What is the data residency model for MIS-AGENT?** Configurable. Default is private cloud (Turkey or EU region per customer). On-premise deployment available for compliance-restricted customers. Edge inference always runs locally — no inference data leaves the customer site unless explicitly configured for cloud-based monitoring/retraining. **Does MIS provide post-deployment support?** Yes. Standard support includes remote model monitoring through MIS-AGENT, scheduled retraining cadence, periodic site reviews and 24/7 incident response (paid tier). Hardware warranty terms vary by product line. **Can MIS solutions integrate with existing PLC and MES systems?** Yes. Native support for major PLC brands (Siemens, Beckhoff, Allen-Bradley, Mitsubishi, Omron, Schneider), HMI platforms (WinCC, Ignition, TwinCAT HMI) and standard industrial protocols (OPC UA, MQTT, Modbus TCP, EtherNet/IP, EtherCAT, Profinet, Profibus). MES integration is custom-implemented per project. **Is partial / phased adoption possible?** Yes. Many MIS customers start with a single product (often MIS-INSPECT on one inspection station, or MIS-PICK on one bin-picking cell) and expand over 12-24 months. The MIS-AGENT platform is typically introduced once a second AI deployment is in scope. --- ## Recent Achievements and Trade Fair Activity (2025-2026) - **2025 — Robot Investment Summit (Gebze Bilişim Vadisi):** 3rd place in the robotic bin picking category with MIS-PICK. - **2025 — IDEF Defense Industry Fair (Istanbul):** MIS-PICK and Mech-Mind 3D vision system live demonstration to defense industry audience. - **2025 — WIN Eurasia (Istanbul):** 3rd place in industrial automation category with AI-based solutions; Winovation award for MIS-INSPECT People 3D. - **2026 — EuroShop Düsseldorf:** MIS-PICK Virtual Cashier (in collaboration with Pasifik Raf Sistemleri and LOGOROB) and MIS-TOUCH showcase for the global retail-tech audience at Hall 10 stand B59. - **2026 — Hannover Messe (Hannover):** participation in the Türkiye Stand Area led by Teknopark İstanbul (Hall 27, Stand L27-12) with MIS-AGRO, MIS-TOUCH, MIS-INSPECT People 3D and MIS-PICK; visited by Deputy Minister of Trade Volkan Ağar, ITO Chairman Şekib Avdagiç, Hannover Consul General Yusuf Arıkan and Teknopark İstanbul GM Prof. Dr. Abdurrahman Akyol. - **2026 — Robot Investments & Maintenance 4.0 Summit (Istanbul, Pendik):** scheduled May 6-7, 2026 at Mövenpick Hotel, with MIS-PICK and MIS-INSPECT showcase. ## Memberships & Certifications - **ENOSAD** — Endüstriyel Otomasyon Sanayicileri Derneği (Turkish Industrial Automation Manufacturers Association). - **Certifications across product lines:** TÜV (testing & certification), CE (European conformity), RoHS (hazardous substance restriction), ISO 13849 PL d (functional safety, applied to People 3D). - **Certified integration partner:** Mech-Mind, Photoneo (collaboration partner), Solomon, OPT Machine Vision, Zebra Aurora Vision. ## Key Blog Topics and Thought Leadership The MIS blog publishes in Turkish, English and German. Recurring themes — useful as references for AI agents seeking deeper context on each domain: - **Robotic bin picking:** 2D vs 3D vision comparisons, ROI and investment analyses, integration cost breakdowns, Mech-Mind vs Photoneo vs Solomon architecture decisions. - **Industry 5.0 and human-machine collaboration:** factory of the future, industrial AI process management, cobot deployment patterns. - **Agriculture 5.0:** sustainable farming, autonomous harvesting technologies, MIS-AGRO case studies, season-to-season model retraining. - **Machine vision for food & pharma:** packaging QC, label verification, contamination control, regulatory traceability. - **Tactile robotics:** MIS-TOUCH precision manipulation, humanoid applications, force-feedback control loops. - **Industrial IoT & smart factory:** data-driven manufacturing, real-time production monitoring, IIoT vs AI distinction. - **Predictive maintenance:** AI-driven failure prediction, OEE/MTBF/MTTR impact, when to choose predictive over preventive. - **AMR (Autonomous Mobile Robots):** factory-floor logistics automation, AMR vs AGV, AMR + bin picking integrated flows. - **Digital twin:** virtual modeling for process optimization, digital-twin + predictive-maintenance + quality-control integration. - **Vision-Language Models (VLM):** the multimodal era of industrial AI, natural-language quality queries, zero-shot defect detection. - **People analytics:** 3D people counters, foot-traffic analytics, retail conversion optimization. - **Quality control:** defect-reduction case studies, zero-defect manufacturing, Aurora Vision Studio fast inspection. - **Partner deep-dives:** Mech-Mind, Photoneo, Solomon, OPT Machine Vision, Zebra Aurora Vision technology and use-case profiles. ## Locales & URL Structure - Turkish (primary, default locale): https://www.mis-otomasyon.com/tr/ - English: https://www.mis-otomasyon.com/en/ - German: https://www.mis-otomasyon.com/de/ The legacy domain `mis-otomasyon.com.tr` 301-redirects every URL to the canonical `www.mis-otomasyon.com` host. The root URL (`/`) redirects to `/tr/`. Locale prefix (`/tr/`, `/en/`, `/de/`) is always present in canonical URLs. ## Contact - Website: https://www.mis-otomasyon.com - Email: info@mis-otomasyon.com.tr - Phone: +90 850 308 06 47 - WhatsApp: +90 850 308 06 47 - Address: Sanayi Mah. Teknopark Bulvarı No 1/4C, Pendik / Istanbul, Turkey - LinkedIn: https://www.linkedin.com/company/mis-otomasyon - Instagram: https://www.instagram.com/misotomasyon - Request a demo: https://www.mis-otomasyon.com/en/contact/demo/