AMR Signals in Shared Workspaces: Make Every Cue Match a Predictable Motion

In manufacturing plants, line-side stores, and warehouse aisles, autonomous mobile robots share space with people, forklifts, and material carts. A vehicle may have obstacle detection, speed control, and stopping functions, yet the people around it still need to understand what it is likely to do next. At an intersection, a pedestrian crossing, a workstation approach, or a brief holding area, the practical question is immediate: will the vehicle proceed, yield, turn, or remain where it is?

Many facilities answer that question with a buzzer or a warning light that is active almost all the time. A robot beeps when it moves, beeps when it sees a person, and may keep beeping while stopped. That approach can turn meaningful information into background noise. A stronger approach treats audible and visual cues as part of robot behavior: when a cue is presented, the vehicle's speed, path, and stopping behavior should support the message; when a person recognizes the cue, the expected response should be clear.

The Objective Is Not a Louder Signal but an Interpretable Motion

Warning cues do not replace protective sensing, speed control, traffic planning, or the application-specific risk assessment. The robot's safety performance still depends on the applicable assessment, the verification of the vehicle and its application, protective-field design, and site operating procedures. ISO 3691-4:2023 specifies safety requirements and verification methods for driverless industrial trucks and their systems. ANSI/A3 R15.08 similarly frames acceptable risk in a specified operating environment as a key basis for industrial mobile robot design and use. Audible and visual signals should make an operating strategy easier to perceive, not transfer the burden of avoidance to a nearby person.

Each cue should therefore answer one question about the current maneuver. A brief indication before entering a shared zone can correspond to a lower speed and a conservative path. A visible state at a blind corner can communicate an intended approach or a yielding condition. A controlled-stop indication can help personnel distinguish a normal wait from a condition that needs trained intervention. When one sound is used for startup, turning, reversing, temporary stopping, and faults, people only learn that a vehicle is somewhere nearby; they do not learn what action is appropriate.

Figure 1: A latent-lifting AMR decelerates before a crossing, pairing a visible vehicle state with a predictable path

Build a Small Vocabulary for High-Value Operating Situations

A useful signaling system does not need to be elaborate. A project team can begin with a small set of situations that matter most to human judgment: normal travel, deceleration toward an intersection, a planned turn, a controlled wait after detecting a person or obstruction, and an abnormal stop requiring intervention. For each situation, document the permitted speed, anticipated path, protective area, and the behavior that people should avoid. Color, flash pattern, sound cadence, and mounting location are communication methods; they should not be designed separately from this motion definition.

For example, a turn indication has value only when the vehicle has actually limited its speed and will not unexpectedly leave its predicted path. A waiting indication should mean a stable stop and a legible clearance boundary, not repeated creeping or probing moves. Lighting conditions, background noise, hearing protection, and the people who occupy the area vary from one facility to another. They must be considered at the actual point of use. General workplace alarm guidance emphasizes that signals need to be perceivable in the ambient light or noise conditions and recognizable by affected workers. That principle is helpful for AMR design, but daily robot cues must not be confused with fire, evacuation, or other emergency signals.

Figure 2: An AMR holds in a controlled waiting condition before a corner while a person remains outside the travel lane

Keep Light, Sound, and Vehicle Behavior on the Same Rhythm

People do not develop confidence in an AMR because it displays a light; they develop confidence because the light and the resulting motion remain consistent over time. If a robot announces that it is about to pass but waits outside an intersection for a long period, or displays a waiting state and then accelerates around an obstacle, observers quickly stop using the signal to predict behavior. In contrast, when an indication appears, the vehicle slows, retains its expected path, and completes a yield or turn in the same controlled manner, people can form a reliable expectation through repeated exposure.

This calls for logic connected to the vehicle's real operating state rather than a generic task name. Instead of sounding throughout a transport mission, a cue can be tied to speed transitions, entry into a defined geographic area, a protective-field event, a confirmed stop, manual control, or a classified fault state. The same signal name may need different motion constraints for latent-lifting AMRs, roller AMRs, autonomous forklifts, and mobile manipulators because their envelope, payload, stopping behavior, and work mechanism differ. A loaded vehicle approaching a turn cannot be treated like an unloaded vehicle on an ordinary travel segment.

Validate Perception During a Real Shift, Not Only at a Desk

A defined color and tone on a drawing do not prove that people will notice or interpret them in the plant. Acceptance and improvement should take place beside noisy equipment, near backlit doorways, at rack-obscured intersections, and in work areas where people may wear hearing protection or focus on material handling. The objective is not maximum volume. It is to confirm that, within a reasonable time and with sufficient predictable space, a person can distinguish the vehicle's direction and intent while the vehicle retains its own ability to slow or stop through its safety functions.

The site should also distinguish normal operating language from conditions that require trained response. A persistently occupied protective area, degraded sensing, a blocked route, or manual takeover is not solved merely by a more urgent buzzer. The fleet control system, local visual management, and personnel procedures need to establish the next action together. Any change to routes, stations, staffing arrangements, or vehicle software should trigger a regression review of the signal plan so that a previously familiar cue does not become ambiguous in a new operating context.

AMR signals are most useful when they support clearer human-robot cooperation rather than add more ambient noise. By tying a limited, memorable set of cues to real deceleration, yielding, turning, and controlled-stop behavior, then validating their perception in working conditions, a facility can help people build stable expectations around mobile automation. For Chengdu Beaver, signal strategy should be designed alongside the vehicle type, task, route, and operating procedures as one reviewable part of a safe industrial mobility system.