Once a site has decided to do something technical about forklifts and people on foot, the next question is what kind of system. The options look similar in a brochure and work very differently on a shift. This guide compares the three main types without naming suppliers, so you can decide on the type before you compare companies.
Every option here is an extra layer, not the first control. HSE's guidance on lift trucks starts with keeping people out:
"Where possible, prohibit pedestrians from areas where lift trucks are operating and only allow access to those who operate truck equipment, or supervisors." (L117, paragraph 121)
After separation, marked routes and signs, it lists further measures, including "visibility aids, for example convex mirrors or CCTV" (L117, paragraph 122). In the US, OSHA's forklift standard puts duties on the driver too: "The driver shall be required to slow down and sound the horn at cross aisles and other locations where vision is obstructed." (29 CFR 1910.178(n)(4))
So the useful question is not which gadget, but which gap in your existing controls you are trying to close. For the basics, see L117 in plain English (UK) or OSHA's forklift rules, 1910.178 (US), and how to check pedestrian segregation.
| Wearable proximity tags | On-vehicle cameras, radar or sensors | Fixed AI cameras on an area | |
|---|---|---|---|
| What it detects | A tagged person near a tagged vehicle | People or objects near a fitted vehicle | People and vehicles in the camera's view, and what they do there |
| Who is warned, and when | Driver and/or pedestrian, in the moment | Driver, in the moment | Usually nobody in the moment: events go to a team to review |
| Does everyone need to wear or carry something? | Yes: every person on foot and every vehicle | No | No |
| Fitting per vehicle | Yes | Yes | No: it watches the area, whatever drives through |
| Covers vehicles you don't own (visiting HGVs, contractors) | Only if they are tagged | Only if they are fitted | Yes, if they are in view |
| Record of patterns over time | Some systems log alerts; depends on the product | Some systems log alerts; depends on the product | Yes: the main purpose |
| Works on existing equipment | No: new tags and readers | No: new kit on each truck | Often, on existing IP cameras |
| Typical ways it falls short | Untagged people, batteries, alert fatigue | Unfitted vehicles, alerts while working close to racking | Only sees what's in view; light, weather and camera position |
| Setup effort | Issue and manage tags; set distances | Fit and maintain each vehicle | Check camera views, set zones and rules |
Products within each type vary, so treat the table as a starting point for your questions, not a verdict on any product.
Being clear about this saves money and disappointment.
Proximity tags only know about people who are wearing one. Visitors, agency staff on their first day and HGV drivers who step out of the cab are exactly the people most at risk, and the most likely to be untagged. Tags need charging or battery changes. In busy areas, frequent alerts can teach people to ignore them, so ask how the warning distance is set and who reviews it.
On-vehicle systems protect the people around the fitted truck, and only that truck. A visiting lorry or a contractor's telehandler has no protection unless it is fitted too. They can warn when the truck is working close to racking or loads, which drivers may learn to tune out. They see around the vehicle, not what is happening at the crossing fifty metres away.
Fixed AI cameras only see what the cameras see. Poor light, rain, glare, a blocked view or a badly placed camera all reduce what they pick up, and areas with no camera need other checks. Most do not warn anyone in the moment: the value is in the record of what happens, for the team to act on, not in stopping the next near miss as it happens.
A rough guide, not a rule:
Yes, and many sites do. The warning in the moment and the record over time do different jobs. A proximity or on-vehicle alert may stop one near miss today. An area record shows that near misses cluster at one crossing at shift change, which is the information you need to move the crossing, change the route or adjust the shift pattern, so fewer warnings are needed at all.
If you already have one system and are adding another, see switching or adding an AI safety camera supplier for how to run both side by side fairly. To trial one or more systems and compare the results, use the pilot scorecard.
RoboK SiteLens is a fixed-camera system. It works with a site's existing IP cameras and recorders over standard RTSP, and supports ONVIF cameras. Where a zone has no usable camera, we can supply one, including our own camera on 4G.
It records near misses between people and vehicles, including forklift and pedestrian near misses, people in vehicle routes, entries to restricted zones you define, PPE, and vehicles over a set speed limit. Each event comes with the time, the zone and the clip. Several detections can run on the same camera, depending on its position.
What it does not do: it does not warn drivers in the cab, and it does not detect crane lifts or suspended loads. It is the record over time, not the warning in the moment. See RoboK SiteLens features for the full list.
Only if visitors are given a tag and wear it. Visiting HGV drivers, contractors and people on their first day are often the hardest to equip, and they are among the people most at risk. If many people on your site come and go, check how tags would be issued, returned and charged, or consider an area system that doesn't depend on what people carry.
Some systems are designed to, often by linking to lights or alarms. RoboK SiteLens does not: it records events for the safety or operations team to review and act on. If you need a warning in the cab, use a proximity or on-vehicle system for that, and an area camera for the record of what keeps happening.
Often not. Many fixed-camera systems can use existing IP cameras, if the camera's position, angle, resolution and lighting are good enough. Ask the supplier to check each camera view and tell you which ones will not work well.
What the rules ask for is that the risk is assessed and controlled. In the UK, HSE's lift-truck guidance (L117) starts with keeping people and trucks apart. In the US, OSHA's forklift standard (29 CFR 1910.178) covers training and safe operation. Whether a warning system or a camera is part of your controls is a decision for your own risk assessment, and it sits alongside separation, training and supervision, not in place of them.
They answer different questions. Tags and on-vehicle systems warn people in the moment. Fixed AI cameras show where and when near misses happen across an area, so you can fix the cause. If you can only start with one, start with the problem you know least about.
To see what a fixed camera records on your own site, start with Safety Mapping: a free two-week trial on one camera, ending in a written report and a review call.
The fairest test of an area camera is one area, two weeks, and a written report of what it saw that you can check against your own records. Safety Mapping: a free two-week trial on one camera.
Last checked November 2026