Project Overview
Motionwell Automation delivered a fully integrated QA laboratory automation system for a leading medical device manufacturer in Singapore. The project (P23078) combined MiR AMR autonomous mobile robots with Universal Robots collaborative robots to create an end-to-end automated testing workflow for quality assurance operations.
The system automates the complete QA testing cycle: sample retrieval from storage, transport to testing stations, automated Instron tensile and puncture testing, result recording, and sample return. All operations are orchestrated by an Allen-Bradley PLC with a custom lightweight WMS (Warehouse Management System) built natively in PLC logic.
System Architecture
AMR + Cobot Compound Robot
The core of the system is a compound mobile robot that combines a MiR AMR base with a Universal Robots cobot arm mounted on top. This configuration enables:
- Autonomous navigation between storage racks and testing stations
- Precise sample pick-and-place at each station
- Flexible routing based on test schedules and station availability
PLC-Native Lightweight WMS
Rather than deploying an external MES or WMS system, Motionwell designed the warehouse management logic directly in the Allen-Bradley PLC. This approach provides:
- 70-position sample storage management across 7 racks
- Real-time inventory tracking with barcode verification
- Task scheduling based on test priorities and station availability
- Server-based work order management for test queue optimization
Instron Integration
The system communicates bidirectionally with multiple Instron universal testing machines:
- Automated zero-reset and speed parameter configuration
- Test start trigger and completion acknowledgment
- Real-time test data monitoring
- Automatic result file naming and network upload
Technical Specifications
| Component | Specification |
|---|---|
| Mobile Robot | MiR AMR with Universal Robots cobot |
| Controller | Allen-Bradley PLC |
| Storage | 70 positions across 7 racks |
| Testing | Multiple Instron stations (tensile/puncture) |
| Traceability | State machine with barcode verification |
| Communication | Ethernet/IP, bidirectional Instron protocol |
Key Engineering Decisions
PLC-native WMS vs. external MES: By building the WMS logic in the PLC, Motionwell eliminated a separate software layer, reducing integration complexity and single points of failure. The PLC handles scheduling, routing, and state management with deterministic timing.
Mixed human/robot operation: The system supports concurrent manual and automated operations. Station safeguarding ensures operator safety while allowing manual sample handling at non-automated stations.
Retry and recovery logic: Verified retry sequences maintain traceability even during error recovery. Operator prompts guide resolution of exceptions without breaking the audit trail.
Technical Details
Three-Layer Closed-Loop Positioning Architecture
The compound robot system achieves reliable sample handling through a three-layer positioning architecture that progressively refines accuracy from macro navigation down to sub-millimeter placement:
-
SLAM Navigation Layer: The MiR AMR uses simultaneous localization and mapping (SLAM) with LiDAR-based environmental scanning for autonomous navigation across the laboratory floor. The SLAM map is updated incrementally as the robot traverses between racks and stations, maintaining positional awareness even as lab furniture or equipment configurations change. Navigation accuracy at this layer is within +/-50mm, sufficient for approaching the target station but not for sample handoff.
-
Cobot Joint Positioning Layer: Once the AMR docks at a station, the Universal Robots cobot arm takes over positioning using its six-axis joint encoders. The cobot operates in a calibrated coordinate frame relative to the AMR’s docking position. Joint-level positioning provides repeatability within +/-0.1mm at the tool center point, enabling the gripper to reach the approximate pick or place location on the rack or Instron fixture.
-
Vision-Based Error Compensation Layer: A wrist-mounted camera on the cobot performs a final alignment step before each pick or place operation. The vision system identifies fiducial markers on the sample trays and test fixtures, calculates the X-Y-theta offset between the expected and actual position, and sends real-time compensation values to the cobot controller. This closed-loop correction compensates for AMR docking variance, rack position drift, and thermal expansion effects, bringing final placement accuracy to within +/-0.5mm.
The three layers operate sequentially on every pick-and-place cycle. If any layer fails its tolerance check, the system retries the current layer up to three times before escalating to an operator alert with the specific failure mode logged.
Server-Based Task Dispatching
A centralized server application manages the work order queue and dispatches AMR transport missions. The server maintains a priority-sorted queue of pending test requests, assigns missions to available AMR units based on proximity and battery level, and tracks mission progress through completion. The server communicates with the Allen-Bradley PLC via Ethernet/IP for real-time synchronization of robot positions and station status.
PLC State Machine for Sample Tracking
The Allen-Bradley PLC implements a formal state machine for every sample position in the system. Each sample transitions through six defined states:
| State | Description |
|---|---|
| Queued | Sample registered in work order, awaiting transport |
| In-Transit | AMR dispatched, sample being transported |
| At-Station | Sample delivered to Instron station, awaiting test start |
| In-Test | Instron test executing, real-time data being recorded |
| Completed | Test finished, results uploaded, sample awaiting return |
| Failed | Test or transport error, operator intervention required |
Every state transition is timestamped with millisecond resolution and logged to the PLC’s internal data table. The timestamp record includes the sample ID, source state, destination state, station ID, and operator ID (if manual intervention occurred). This state history forms the core of the audit trail and can be exported for regulatory review.
The state machine enforces strict transition rules – a sample cannot move from Queued to In-Test without passing through In-Transit and At-Station. Invalid transition attempts are blocked and logged as exceptions.
Related
See how Motionwell serves the laboratory & QA industry and medical device industry.