Autonomous mobile robots (AMRs) can navigate around obstacles and recalculate routes, offering flexibility that differs from the fixed paths traditionally associated with automated guided vehicles. But that flexibility does not automatically resolve congestion, right-of-way conflicts or interactions with people and other vehicles. Clearly defined AMR movement rules—and a practical way to test them—can help manufacturers anticipate how robots will behave in a busy facility.
A research methodology published in 2019 explored this challenge by developing an AMR “Highway Code,” building simulation models and comparing simulated behavior with physical shop-floor observations. Its central lesson for automation teams is that neither simulation nor physical trials alone provide a complete picture. Combining them can expose behavioral differences and help inform deployment decisions.
Why AMR movement rules matter
AMRs use sensor feedback to perceive their surroundings and plan routes around obstacles. In a dynamic manufacturing environment, however, an obstacle may be a pedestrian, a moving load or another vehicle—not simply a fixed object. These participants may approach from different directions, change course or block the space an AMR would use to pass.
Without shared expectations for movement, a robot may stop while another vehicle waits behind it, or several participants may attempt to use the same narrow route. Such interactions can create delays and, in some circumstances, potential collision risks. Rules covering issues such as obstacle response, passing, routing and traffic priority can give system designers and facility teams a common basis for evaluating these situations.
The Highway Code concept brings rules that are often studied separately into a structured framework. The work grouped rules into four broad areas:
- Traffic regulation: coordination and traffic-control approaches for vehicle movement.
- Dispatchment: decisions about pickup, dispatch and delivery.
- Load selection: priorities for choosing which loads or tasks to handle.
- Routing method: approaches for planning and optimizing routes.
This structure is useful because AMR performance on a factory floor depends on more than route planning. Dispatch priorities and traffic behavior can affect whether a planned route remains workable once other activities are considered.
Combining simulation with shop-floor evidence
Simulation makes it possible to examine scenarios that could be time-consuming or hazardous to test repeatedly with physical equipment. Yet a simulation based only on idealized assumptions may miss the details of actual interactions. Conversely, relying solely on physical tests can make it difficult to explore a wide range of conditions efficiently.
The described methodology addresses this gap through an iterative process. First, the team documents relevant features of the manufacturing environment, including paths, junctions, pedestrians, obstacles, vehicle dimensions and operating zones. It then constructs a simulation model and compares its behavior with observations of a physical AMR. Differences between the two can be used to adjust the model, making it a closer representation of the real system.
In the study, researchers compared three motion-planning algorithms—Rapidly exploring Random Tree (RRT), Probabilistic RoadMap (PRM) and KPIECE—against recorded paths of a physical AMR. They assessed differences in both position and direction of travel. RRT showed the smallest average discrepancy in those comparisons and was selected for simulating the observed AMR behavior. The result is specific to the tested scenario and model; it should not be treated as a universal ranking of planning algorithms.
Testing the difference between ideal and real behavior
The study distinguished an “ideal AMR,” modeled to follow a defined set of movement behaviors, from a “real AMR,” represented by physical testing and a simulation adjusted to reflect observed behavior. The idealized model included responses such as waiting when an overtake was unsafe, allowing a faster vehicle to pass, reversing when an obstacle approached too closely from the front, and canceling an overtaking plan if the original route became clear.
Physical testing revealed that the tested robot did not exhibit every behavior in the ideal model. For example, it did not respond to an approaching obstacle from behind unless that obstacle entered its safety field. It also did not reverse when faced with an oncoming obstacle; instead, it replanned based on the environment it detected. When an obstacle was removed during an overtaking maneuver, the robot retained the area in its map and did not immediately generate a new route if its current path remained valid.
These observations illustrate why a robot’s configured behavior should be evaluated in context. The work noted that a commercially deployed system may prioritize completing its mission and may not distinguish pedestrians from vehicles or infer their intentions. A fleet manager or other system with broader information may play a role in coordinating multiple vehicles, but the specific approach depends on the installation.
What scenario-based testing can reveal
Researchers compared ideal and real AMR behavior across five shop-floor situations: a blocked path, blocked overtaking, simultaneous overtaking, a moving obstacle obscured by a corner, and multiple moving obstacles. The scenarios highlight how small differences in movement rules can affect surrounding traffic.
- Blocked path: A load and an AMR meet where passing space is limited. In the modeled real behavior, the load had to reverse to a wider area before the AMR could proceed.
- Blocked overtaking: A faster vehicle approaches from behind. If the AMR does not yield, the following vehicle may be unable to pass cleanly.
- Blind corner: A vehicle appears from around a corner. The ideal model could reverse when the approaching vehicle continued forward, while the tested real behavior did not include that response.
- Multiple moving obstacles: Several vehicles meet an AMR where an alternate pedestrian path is unavailable, potentially leaving traffic waiting or requiring vehicles to reverse.
These are not simply route-planning edge cases. They involve how the robot, people and other vehicles share space—and whether each participant can make room for the others.
Applying the findings to AMR deployment
For engineers and system integrators, the approach suggests several practical steps. Document the routes and interactions that matter in the intended operating area. Identify likely conflicts, especially at narrow passages, junctions and blind corners. Define the movement behaviors the application requires, then test those behaviors in simulation and compare the results with physical observations.
When testing reveals a mismatch, teams can consider whether the response should come from changes to the AMR, its simulation model or shop-floor protocols. The study’s scenario comparisons showed that the absence of particular behaviors could contribute to delays and potential collisions. They also demonstrated how a calibrated simulation could help identify such concerns before wider implementation.
Conclusion
Reliable AMR integration requires more than selecting a navigation algorithm. Manufacturers also need to understand how mobile robots behave when routes are obstructed and traffic participants interact. A structured set of AMR movement rules, combined with observation-based simulation and targeted physical testing, offers a way to evaluate those interactions systematically. For plant teams, the practical value is earlier visibility into potential traffic conflicts—and a clearer basis for deciding whether to adjust robot behavior, facility procedures or both.
Access Official Technical PDF