AMRs and safety: how autonomous mobile robots reduce workplace injuries

Between 2002 and 2020 Italy's Infor.MO surveillance system recorded 245 fatal accidents in which a forklift played an active role, out of roughly 5,500 cases collected. The fact sheet INAIL devoted to them in 2023 describes a recurring scene. In most cases a worker on foot is struck by a forklift. The causes include transit routes for vehicles and pedestrians that were not defined or not separated, and company rules that were not followed. In forklift cases the "third-party activity" factor accounts for 23%, against about 10% across all the accidents analyzed, because the driver becomes the third party who makes the wrong move. An autonomous mobile robot acts on exactly that point, taking the driver out of the sequence and handing surveillance of the space to sensors that don't get distracted. Provided the project is done properly.

Two families of injuries, one common factor

In intralogistics the risks fall into two broad families. The first is musculoskeletal disorders from manual handling: lifting, pushing, pulling and carrying loads cause cumulative injuries to backs, shoulders and limbs, and EU-OSHA ranks them among the most common work-related health problems in Europe. The second is forklift incidents, with pedestrian-vehicle collisions, tip-overs, falling loads and crushing while reversing. They are rarer but far more serious, as the INAIL figures above show.

The common factor is a person standing at the point of risk: the body doing the lifting, someone walking next to a moving truck, the driver exposed to a tip-over. The hierarchy of controls puts eliminating the hazard ahead of protecting against it, which is why reducing exposure is worth more than any warning sign.

What a mobile robot removes, and what it adds

On the first family the effect is direct. An AMR that carries pallets, totes and roll cages between two fixed points cuts lifting, pulling and pushing, and the operator moves to supervision or value-adding work, often in a better posture. On the second the mechanism is different. The robot doesn't get distracted, doesn't force maneuvers and doesn't reverse without monitoring. It travels at controlled speed and carries safety laser scanners that slow or stop the vehicle when a person enters the monitored area, before contact.

It would still be wrong to describe it as risk-free. It brings hazards of its own: sudden stops in an aisle where a manned truck is following, people who after a few weeks trust it too much and cut across its path, and charging stations and pick-up and drop-off points where distances shrink. AMR safety is therefore a property of the installation before it is a property of the vehicle. It depends on how the functional safety system (sensors, control logic, protective fields) has been sized and verified for that specific area. How AGVs and AMRs behave differently when something blocks the way is covered in our article on AGVs and autonomous mobile robots.

Laser scanners and protective fields: how they are sized

The core of AMR safety is perceiving the space around it. Safety laser scanners generate configurable two-dimensional fields around the vehicle, usually on two levels. The outer warning field makes the robot slow down and signal when a pedestrian enters. The inner protective field brings it to a full stop before contact if the person comes closer. The fields change with speed and direction of travel, widening at higher speeds and in the direction of motion, because stopping distance grows with speed and with the mass being carried.

Sizing therefore starts from the real stopping distance of the fully loaded vehicle, measured on the actual floor, not from the theoretical unloaded value. The scanners belong to the family of electro-sensitive protective equipment covered by the IEC 61496 series. For devices that sense by diffuse reflection, such as laser scanners, the specific part is IEC 61496-3, now in its fourth edition from 2025, which replaced the 2018 edition. The reliability of the safety-related control chain, from sensor to brake actuator, is designed with the Performance Levels of ISO 13849-1:2023, and the level each function requires follows from the risk assessment.

The standards to write into the specification

In the specification the sequence is linear. Start from the risk assessment under ISO 12100, apply the specific requirements of ISO 3691-4 for driverless trucks, and design the safety functions with the reliability ISO 13849-1 calls for. The devices must comply with IEC 61496-3, and everything is documented for CE marking. Until January 19, 2027 the reference remains the Machinery Directive 2006/42/EC, under which EN ISO 3691-4:2023 gives presumption of conformity since its citation in the Official Journal by Implementing Decision (EU) 2024/1329. From January 20, 2027 Regulation (EU) 2023/1230 applies, replacing the Directive and adding explicit requirements for the supervision of autonomous mobile machinery. A project assessed today and commissioned in 2027 should state in the contract under which regime the declaration of conformity will be issued and who signs it for the whole system. Responsibility in fleets that mix brands is developed in our article on automating pallet handling in a running warehouse.

Shared areas are designed before the vehicles are bought

Unlike many traditional AGVs, AMRs are designed to work in areas shared with people, without segregated routes. That is an operational advantage, but it moves the center of gravity of safety onto the environment. Speeds have to be set zone by zone according to pedestrian density, and intersections need written right-of-way rules and, where necessary, signals. Floor markings, signage and lighting deserve the same care as for manned trucks, and breakdowns, persistent obstacles and blockages need written procedures before start-up. Where possible, the fastest flows should be kept apart from the busiest pedestrian zones. It is the same advice the INAIL fact sheet gives for manned forklifts, with physical separation and clearly marked crossings.

Near misses as a data source

Technology reduces risk but does not replace a safety culture. AMRs do, however, produce data that manned trucks don't: safety stops, slow-downs, detected obstacles, each with a place and a time. Treating them as near misses to analyze, rather than as nuisances to silence, makes it possible to find the critical points of the layout and risky behavior before they turn into an injury. The same INAIL fact sheet recommends a procedure for recording forklift-related near misses. With an automated fleet that record comes ready-made, and its periodic review belongs in the operating plan. The economic weight of prevention then enters the business case, as explained in our analysis of AMR ROI, TCO and payback.

What to send us for an assessment

A first assessment of shared areas and safety requirements needs the following:

• a dimensioned floor plan with aisles, intersections, doors, gates and pedestrian zones;
• pedestrian and manned-truck flows, with the busiest time slots;
• loads handled, weights, pick-up heights and current vehicle speeds;
• floor condition, slopes and ramps along the routes;
• handling-related injuries and near misses in recent years, if recorded;
• the risk assessment for the area concerned, or its relevant parts.

You can send them by email to info@pitech-solution.com or on WhatsApp. The available technologies are described on the pages about robotic forklifts and AMRs for pallet handling and industrial automation and material handling robotics. When an arm rides on the vehicle as well, more standards come into play, as our article on mobile manipulators explains.