The fastest-growing corner of forklift safety is the one with nobody on the truck. Automated guided vehicles (AGVs) and autonomous forklifts remove the operator error that drives most forklift accidents, but they move the risk around instead of removing it: a quiet machine with no driver, pedestrians who forget there is no one to see them, and a safety case that treats the entire facility as part of the system. This guide explains how driverless forklift safety is actually engineered and governed, and sells none of these systems.
What AGVs and autonomous forklifts are
An AGV forklift combines a conventional forklift’s load handling with automated navigation. The truck follows a route set by laser scanning, magnetic tape, QR codes, or vision systems, and executes tasks such as pallet handling, stacking, and loading without an operator, as the explainer from AiTEN Robotics describes. The same capability ships across the truck range: automated guided carts and tow tractors, counterbalance trucks, reach trucks, pallet stackers, and very narrow aisle trucks, the catalog Toyota and Linde publish. Navigation splits into two families, per Linde: track-guided, using magnetic or visual aids that are cheap to lay but costly to change, and natural feature navigation, where laser sensors map the environment and routes can be adjusted with little rework.
Terminology matters. An AGV follows fixed paths and stops when it meets an obstacle, requiring intervention. An autonomous mobile robot (AMR) plans its own path in real time and routes around obstacles. Load capability differs in the usual case too: AGVs commonly move loads up to several tonnes along fixed routes, while AMRs typically handle under a tonne in dynamic settings, AiTEN notes, with the two converging as AMR-style navigation appears on AGV platforms.
The standards that govern driverless safety
Two documents define what safe means. Internationally, ISO 3691-4, Industrial trucks, safety requirements, part 4, driverless industrial trucks and their systems, applies to manufacturers and operators. First published in 2020 and updated in 2023, it defines the performance levels of the safety functions, person detection, modes of operation, and the braking system, and it describes the risk-reduction and validation procedure for the final installed environment, per safety-automation firm Pilz. The 2023 update sharpened the active detection field and the operational stop, the stop function and emergency stop spacing, and the conditions for running at higher speed in restricted zones, which requires a documented hazard analysis. The standard runs about 82 pages, as the integration guide from JL Robotics summarizes, and its safety-function reliability framework aligns with ISO 13849-1.
In North America, ANSI/ITSDF B56.5, Safety standard for guided industrial vehicles, is the reference, first published in 1978 and revised to the 2024 edition. B56.5 takes a system-level view: safety emerges from the interaction of the vehicle, its control system, the facility, and the people in it, so the standard assigns responsibilities across the manufacturer, the system supplier, and the operator, and covers automatic braking, speed control, collision avoidance, pathway design, warning devices, and emergency stop mechanisms, as the compliance guide from Reeman Robotics and the ANSI blog explain. It also addresses vehicles converted from manned to unmanned operation.
What the safety features actually are
Across both standards, the practical safety case is consistent, per Reeman’s comparison. Obstacle detection and collision avoidance use LiDAR, ultrasonic, or camera sensing that stops the vehicle before contact. Emergency stop systems bring the truck to an immediate halt and require a manual restart. Speed and travel path controls reduce speed automatically in high-traffic zones. And audible and visual warning devices tell nearby workers a truck is moving. The facility side carries the rest of the load: travel paths marked with barriers, floor markings, and electronic zone controls, detection zones validated at maximum speed and maximum payload, trained personnel who know the warnings and exclusion zones, scheduled maintenance of sensors and brakes, and a re-run of the risk assessment whenever routes or layouts change.
The regulatory reality in the United States
There is no dedicated federal rule for autonomous forklifts. 29 CFR 1910.178 governs powered industrial trucks generally, the standard behind our plain-English guide to 1910.178, and OSHA enforces the general duty clause where a recognized hazard is left unaddressed. The consensus standards are voluntary, but they are the yardstick: a facility that deploys AGVs without the documented hazard analysis and the safety functions ISO 3691-4 and B56.5 describe has a weak answer when an incident happens. State-plan states may layer on their own requirements, and your state’s injury series provides the local denominator for the budget meeting.
Coexistence with manned trucks and pedestrians
AGVs increasingly share warehouses with human-driven forklifts, and the manufacturer guidance is consistent: proper layout planning, safety zones, and integration strategies are essential for mixed operation, as Toyota puts it. The risk profile shifts toward pedestrians and maintenance staff, because a driverless truck is quiet, predictable in its lane, and dangerous precisely when someone assumes it will behave like a manned forklift and yield. That is the same struck-by exposure mapped in the warehouse pedestrian safety guide and the forklift blind spots guide, now with no operator in the loop to see you at the last moment. Automation is a capital project, not a substitute for a pedestrian program, and it layers with the detection families on the systems index the same way AI camera systems do on manned trucks.
The data picture
The search data for this guide shows the demand is real and early: ‘agv forklift’ runs at roughly 480 monthly US searches with negligible keyword difficulty, and ‘driverless forklift’ near 90, with commercial intent, per the DataForSEO snapshot for this guide. That is a buying audience arriving before the standards education is common, which is the gap this guide exists to close. The same honesty this site applies everywhere, on the methodology page, applies to automation: the machine is not the safety program, the analysis and the zones are.
If you are evaluating automation, start the way every evaluation here starts: score your current exposure with the Risk Scorer, understand your pedestrian and intersection risk before you add a machine that never blinks, compare the systems index for the human-driven fleet that will share the floor, and when the scope is pinned, get comparable numbers through the quote form.