Walk into any modern automated facility and you'll see the same challenge playing out in a hundred different configurations: people and machines sharing space. Robots load pallets, AMRs cross aisles, conveyors feed presses, while technicians, operators, and forklift drivers move through all of it. Keeping those humans safe isn't a single-product decision. It's a system design problem, and it's worth thinking about the way you'd think about any other engineering architecture: in layers, with defined functions, measurable performance, and no single point of failure.
This post walks through the major building blocks of a modern machine safety system, safety-rated LiDAR, light curtains, 3D stereo sensing, interlocked guarding, and safety controllers, and how they fit together into a coherent design.
Start With the Standards, Not the Hardware
Before selecting any component, define the required performance of each safety function. Two frameworks dominate:
- Performance Level (PL) under ISO 13849-1, rated PLa through PLe
- Safety Integrity Level (SIL) under IEC 61508 / IEC 62061, rated SIL 1 through SIL 3
A risk assessment, severity of injury, frequency of exposure, possibility of avoidance, will provide the required PL or SIL for each hazard. Only then do you select hardware that meets or exceeds it. A perimeter guarding function on a high-energy press might demand PLe / SIL 3, while a slow-moving cart's field monitoring might be adequately covered at PLd / SIL 2. Over-specifying everything wastes budget; under-specifying anything creates liability.
Third-party certification matters here. Look for devices certified by bodies like TÜV, and for compliance with the standards specific to the device category, ISO 13849-1 for control systems, ISO 14119 for interlocking devices, IEC 61496 for electro-sensitive protective equipment.
Layer 1: Area Scanning With Safety LiDAR
Safety-rated 2D LiDAR scanners are the workhorse of area protection, both on mobile platforms and around fixed cells. A typical unit scans roughly a 270° field with a protective range of several meters, resolving objects small enough to reliably detect a human entering the zone.
The design power of safety LiDAR comes from configurable fields. Rather than a fixed detection envelope, you define warning zones and protective zones in software — fan-shaped, rectangular, or irregular polygons that match the actual hazard geometry. A common pattern:
- Warning field (outer): slows the machine or triggers an audible alert
- Protective field (inner): initiates a safety-rated stop
On AMRs and AGVs, these fields typically switch dynamically with speed and steering angle. A vehicle turning left at full speed needs a very different protective envelope than one moving forward slowly. Verify that your safety device/scanner supports enough field sets and fast enough switching for your motion profile.
Two other capabilities worth requiring: self-diagnosis with real-time fault reporting (a safety sensor that fails silently is worse than no sensor, because it creates false confidence), and standard industrial interfaces; Ethernet, CAN, safety-rated I/O; so the scanner integrates cleanly with the PLC or robot controller rather than through fragile adapter logic.
Layer 2: Point-of-Access Protection With Light Curtains
Where people intentionally reach into or walk through an opening, a load station, a pass-through window, safety light curtains provide the defined plane of protection. Key specification points:
- Type 4 designation (per IEC 61496) for the highest-risk applications, achieving PLe / SIL 3
- Resolution matched to the body part: finger protection (~14 mm), hand protection (~30 mm), or body detection for perimeter use
- Blind-zone-free, edge-to-edge designs where mounting space is tight, dead zones at the ends of a curtain are a classic source of undetected intrusion
- Dual-CPU, dual-output self-testing architectures, so the failure of a single circuit can't silently disable the protective function
- Optical or line synchronization options to prevent mutual interference when multiple curtains operate near each other
Remember that a light curtain's safety distance calculation (per ISO 13855) depends on resolution and response time. A faster, finer curtain can be mounted closer to the hazard, which often matters more for cell footprint than the sensor's sticker price.
Layer 3: 3D Perception and AI-Based Human Detection
2D scanning covers a plane; humans exist in three dimensional space . Stereo vision safety sensors add volumetric awareness: a 3D point cloud over a wide field of view (on the order of 120° × 50°) at working distances up to tens of meters, with AI-based human recognition distinguishing a person from a pallet or a passing vehicle.
This category is newer than LiDAR and light curtains, and design maturity varies, so evaluate carefully:
- Laser Class 1 emission, safe for continuous exposure in occupied spaces
- Mechanical robustness: shock and vibration resistance if mounted on moving equipment
- Environmental rating: IP65 minimum for washdown-adjacent or dusty environments; IP67 where needed
- Complementary triggering: some architectures use LiDAR detection to trigger real-time recording or classification from the 3D sensor, giving you both a certified stop function and rich contextual data for incident review
The practical role of 3D sensing today is often supplementary, enhancing situational awareness, enabling smarter mute/override logic, and providing forensic data, while the certified stop function still runs on established safety-rated devices. Assign each device the role its certification actually supports.
Layer 4: Guarding and Interlocks
Physical guarding remains the most reliable protection there is, a fence doesn't have a response time. But every fence needs doors, and every door needs an interlock. Modern interlocking options include:
Electromagnetic locking switches with guard locking forces up to ~1300 N, keeping a door physically closed until the hazardous motion has stopped. Look for PLe / SIL 3, Type 4 ratings, RFID-based coding (with high-level “unique” coding available for anti-tampering), and smooth end-face designs that tolerate door sag and installation misalignment without jamming.
Non-contact safety switches where hygiene, wash-down, or alignment tolerance make mechanical keys impractical.
Cascading support deserves attention in larger cells: some switch families allow 20–30 units to be daisy-chained on unified safety outputs, dramatically reducing safety input count at the controller. Just verify that cascading preserves the required PL, series-connecting mechanical contacts can mask faults and degrade diagnostic coverage if not designed for it.
Status indication is more than convenience. Multi-angle indicator lights that show door and lock state at a glance reduce the operator confusion that leads to defeat attempts, and defeated interlocks are one of the most common root causes in machinery injury investigations. ISO 14119 explicitly requires designing against foreseeable defeat.
The Integration Layer: Safety Controllers
All of these devices terminate somewhere, and that somewhere should be a certified safety controller — either a configurable safety relay/controller or a full safety PLC, rated to PLe / SIL 3.
Modern compact controllers (in slim 22.5 mm form factors) accept multiple device types on one platform: emergency stops, interlock switches, light curtains, scanner OSSD outputs. The design questions to ask:
- Does the controller's logic capacity cover your zone structure, including muting, blanking, and mode-dependent behavior?
- Can it maintain dual-channel monitoring and cross-fault detection across every input?
- Does its response time, added to sensor response and machine stop time, still satisfy your ISO 13855 safety distances?
- Is the configuration tool auditable, so the safety logic can be reviewed and version-controlled like any other engineering artifact?
Putting It Together
A defensible safety architecture usually ends up looking like concentric rings:
- Awareness: Warning fields, indicator lights, audible alerts
- Detection: LiDAR protective fields, light curtains, 3D human recognition
- Physical prevention: Guarding with locked, monitored access points
- Certified logic: A safety controller executing the stop functions with verified integrity
No single layer carries the whole burden, and the failure of any one device degrades gracefully rather than catastrophically. That's the goal: not a collection of safety products, but a safety system, specified from the risk assessment, built from certified components, integrated through auditable logic, and validated end to end.
The hardware landscape keeps improving, better resolution, smarter perception, easier integration. But the engineering discipline stays constant: define the required performance first, design in layers, verify the numbers, and never let a data sheet substitute for a risk assessment.
Working through a safety architecture for a new cell or mobile platform?We spend our days helping engineering teams match sensing hardware to the performance level their risk assessment actually requires, often at the same safety class as the incumbent options, at a meaningfully lower cost. If you'd like a second set of eyes on your zone layout or component selection, book a conversation with one of our engineers . No pitch, just an engineering conversation.
Morpheus Tek, Inc. Sight at the Edge Blog morpheustek.com Sales@morpheustek.com +1 (302) 789-0421