# Shadowing method cuts multi-robot transport messages by over 93% versus Leader-Follower in simulation tests

- Published: 2026-11-16 23:49 UTC
- Section: Machines & Making › Robots
- Tags: multi-robot cooperative transport, virtual geometry shadowing, formation control, Leader-Follower, centralized virtual structure, event-triggered communication, Anytime D*, Gazebo simulation, smart factory robots, communication reduction
- URL: https://realtech.media/news/rt-777dd9981f0cbcf734d24173_shadowing_method_cuts/

**Hoolark Jung, Donghyun Lee and Hyunchul Tae proposed Shadowing: robots follow a pre-sent trajectory and transmit only on events. In 16,200 trials per planner, there were no formation collapses.**

![Shadowing method cuts multi-robot transport messages by over 93% versus Leader-Follower in simulation tests](https://realtech.media/media/2026/11/2d12054938415f43.jpg)
*Ckalafatis24, Wikimedia Commons, CC BY-SA 4.0*

A research group of Hoolark Jung, Donghyun Lee and Hyunchul Tae has proposed Shadowing, a formation control method for multi-robot cooperative transport. According to the paper published via doi.org, it reduced the number of inter-robot messages by more than 93% relative to the Leader-Follower scheme and by more than 96% relative to the centralized virtual structure (Centralized-VS) [1].

Existing approaches depend on continuous communication. Leader-Follower schemes require constant exchange between robots, and centralized virtual structure approaches need continuous position feedback. As fleets grow in logistics automation and smart factories, such control-rate traffic loads the shared wireless channel and creates computational bottlenecks, which the authors link to risks for transporting fragile payloads [1].

## How Shadowing works

The robots follow a pre-transmitted time-space trajectory of a virtual polygon framework. Each robot keeps its assigned position independently, using pre-stored timestamps, so the formation and the transport continue during communication outages [1].

Communication consists of one initial trajectory broadcast plus event-triggered broadcasts. These are sent when a dynamic obstacle is detected, for replanning, and when the position error exceeds the threshold ε_max, for realignment [1].

## Test results

The quantitative campaign comprised 16,200 trials per path planner: 3 map types, 4 obstacle densities, 3 formation methods and 3 formation sizes (2, 4 and 6 robots), 150 trials per condition. Each dynamic obstacle appeared at a random time uniformly distributed over 1–2 s and stayed static afterwards. Continuously moving obstacles were evaluated separately in a physics-based Gazebo study [1].

With Anytime D* as the planner, message count fell by 93.8%, 94.3% and 95.5% relative to Leader-Follower for 2, 4 and 6 robots, and by 97.8%, 97.0% and 97.4% relative to Centralized-VS. Missions were completed 25.0%, 21.2% and 21.5% faster than with Leader-Follower. Compared with Centralized-VS, mission time was 4.6–10.0% longer, while formation error was 42–48% lower [1].

With the A* planner, message reductions relative to Leader-Follower were 93.7%, 93.7% and 94.8%, and relative to Centralized-VS 97.8%, 96.8% and 97.0%. Missions were 19.4–23.9% faster than with Leader-Follower, and formation error was 45–51% lower than with Centralized-VS [1].

The absolute message gap grows roughly linearly with the number of robots, because the broadcast count depends on the event rate rather than on the control frequency. The authors recorded zero mission failures and zero formation collapses under both planners in all conditions. The reduction figures refer to message counts; the paper also reports a complementary data-volume analysis [1].

## Sources

1. [doi.org](https://doi.org/10.1007/s10846-026-02488-5)
2. [bioengineer.org](https://bioengineer.org/virtual-geometry-shadowing-cuts-multi-robot-transport-communication-by-over-ninety-three-percent)

---

Source: RealTech, realtech.media — https://realtech.media/news/rt-777dd9981f0cbcf734d24173_shadowing_method_cuts/
