Forklifts and Pedestrians Sharing the Floor: How to Prevent Collisions with Separation and Flagman AI Detection
Vehicles were involved in 66% of Australian worker deaths in 2024. What work health and safety law requires for forklift and pedestrian safety, and where AI detection genuinely helps.
In brief
- Vehicles were involved in two out of every three Australian worker deaths in 2024, and forklifts remain one of the most common sources of serious harm in warehouses and factories.
- Work health and safety law requires businesses to stop powered mobile plant from colliding with people, and to choose controls in a set order, starting with separating people from vehicles.
- AI pedestrian detection is a valuable extra layer that warns operators and, on suitable forklifts, can slow or stop the truck. It strengthens separation and traffic management. It does not replace them.
- The sites that get real value from technology are the ones that measure a baseline first, fix the layout, and then add detection where people and forklifts still have to share space.
Every safety manager has heard the story. A forklift reverses out of a racking aisle, a picker steps out from behind a pallet at the same moment, and the only thing that prevents a serious injury is luck. Most of these near misses are never reported. The ones that are often end up as a line in an incident register and a reminder at the next toolbox talk.
That approach is no longer good enough, legally or commercially. Australian regulators expect businesses to treat forklift and pedestrian interaction as a high-consequence risk and to manage it with engineered controls, not good intentions. This article sets out what the law requires, how the risk shows up on real sites, and where AI-based detection fits into a credible control plan.
The scale of the problem in Australia
Safe Work Australia’s latest national figures show that 188 workers were killed at work in 2024. At least one vehicle was involved in 66% of those deaths, which is 124 people. Vehicles, in other words, are not one hazard among many. They are the single largest factor in Australian workplace fatalities.
Manufacturing recorded 7 worker deaths in 2024 and 14,800 serious workers compensation claims in 2023 to 2024. Behind each claim sits lost time, investigation costs, disrupted production and, too often, a worker whose life has changed permanently.
Forklifts are a large part of that picture. They are heavy, often carry loads that block the operator’s view, turn within their own length and spend much of the day reversing. Put them in a busy warehouse with pickers, contractors, delivery drivers and visitors, and the conditions for a collision are present every shift.
What the law requires
In every state and territory except Victoria, the model Work Health and Safety Regulations set out specific duties for powered mobile plant. Regulation 215 requires controls so that mobile plant does not collide with pedestrians or other mobile plant, and requires a warning device where there is a risk of collision. Regulation 218 adds further requirements for industrial lift trucks such as forklifts.
Victoria operates under its own Occupational Health and Safety Act 2004 and Occupational Health and Safety Regulations 2017. The wording differs, but the expectation is the same: employers must identify the risk of plant colliding with people and control it so far as is reasonably practicable.
The law also tells you how to choose controls. Regulation 36 sets out the hierarchy of controls:
- Eliminate the risk where you can, for example by removing forklift traffic from pedestrian areas entirely.
- Substitute, isolate or engineer the risk down, for example with physical barriers, dedicated walkways, interlocked gates and separate loading zones.
- Use administrative controls, such as traffic management plans, speed limits, exclusion zones and training.
- Use personal protective equipment, such as high visibility clothing, as the last line of defence.
Regulators look closely at this order. A site that relies on high visibility vests and a toolbox talk, while forklifts and pedestrians freely share the same aisles, will struggle to show it has done what is reasonably practicable.
Why traditional controls still leave a gap
Most Australian warehouses and factories already have a traffic management plan, marked walkways and trained operators. So why do collisions and near misses continue?
The honest answer is that the workplace changes faster than the plan. Racking moves, temporary storage blocks a walkway, a new contractor does not know the rules, or a peak period pushes people to take shortcuts. Mirrors and reversing beepers help, but they still depend on a tired operator noticing a person who has stepped into a blind spot at exactly the wrong moment.
The solution: how Flagman detects pedestrians in forklift blind spots
AI pedestrian detection adds a layer of awareness that does not get tired or distracted. Cameras mounted on the forklift cover the blind spots at the rear and sides. An edge computer on the truck analyses the video in real time and recognises when a person enters a defined zone.
In Flagman’s standard forklift configuration, a warning sounds when a person comes within about 2.5 metres of the truck, and a critical alert is triggered at about 1.5 metres. Alerts can be audible, visual and delivered in local languages, so operators understand them instantly in noisy environments.
On forklifts whose control systems accept an external signal, mainly modern electric models, the system can also slow or stop the truck. On older or internal combustion forklifts, it can usually provide only the warning. This depends entirely on the forklift manufacturer’s controls, so it must be confirmed truck by truck during a site assessment.
Because processing happens on the device, the system keeps working without an internet connection. Events can be recorded and reviewed on a central dashboard, which turns near misses that would otherwise go unreported into data that safety teams can act on.
Understanding the limitations
Any credible safety technology should come with a clear statement of what it cannot do. AI pedestrian detection is no exception.
- It does not replace separation. Barriers, walkways and exclusion zones remain the primary control.
- It depends on what the cameras can see. Heavy dust, glare, very low light or a dirty or blocked lens can reduce detection, so cameras need cleaning and routine checks.
- Alarm fatigue is real. Zones that are set too wide create false alerts, and operators learn to ignore them. Zones must be tuned to each truck and each area.
- Automatic slowing or stopping is a safety function. Before it is relied on, it needs proper engineering assessment and testing on the specific truck and site.
Being upfront about these limits is not a weakness. It is what allows a safety manager to put the technology in the right place in the control plan and defend that decision to a regulator, a board or an insurer.
A practical rollout plan for your site
- Map the interaction points. Walk the site and mark every place where people and forklifts share space: loading docks, racking aisles, doorways, battery charging areas and amenities.
- Fix the layout first. Install barriers, mark walkways and create controlled crossing points before you buy any technology. These are cheaper and more reliable than any sensor.
- Measure a baseline. Record near misses and close calls for two to four weeks. Without a baseline, you cannot prove whether any later control has worked.
- Target detection where separation is not possible. Start with the trucks and zones that carry the most interaction, rather than equipping the whole fleet at once.
- Consult your people. Work health and safety law requires consultation before significant changes, and operators know exactly where the real blind spots are.
- Review monthly. Compare alert and near-miss data against the baseline with your safety committee, and adjust zones, layout or training based on what the data shows.
The outcome you can measure
When separation, traffic management and Flagman work together, sites can expect three outcomes they can measure on their own floor. Operators receive a warning at the moment a person enters a blind spot, rather than relying on mirrors alone. Near misses that previously went unreported become a time-stamped record, showing exactly which trucks, aisles and shifts carry the most risk. And because every engagement starts with a baseline, the site can compare interaction data before and after changes to layout, procedures or training, giving managers and boards evidence that controls are working.
Key takeaways
Forklift and pedestrian collisions are one of the most predictable serious risks in Australian industry, and the law is clear about how they should be managed. Start with separation and good traffic management. Then use AI detection to close the gaps that human attention and static controls will always leave, and measure the results on your own floor.
Next step
Sharktech Global is the Australian delivery partner for Flagman and is working with a limited number of founding sites in New South Wales. Each engagement begins with a baseline safety assessment, so results are measured on your site rather than promised in a brochure. Learn about Flagman in Australia or request a safety assessment.
Sources: Safe Work Australia, Key Work Health and Safety Statistics Australia 2025; model Work Health and Safety Regulations 2011, regulations 36, 215 and 218; Occupational Health and Safety Act 2004 (Vic). This article is general information, not legal or safety advice for your site.