Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees

arXiv:2603.10245 · eess.SY, cs.SY · Submitted 2026-03-10 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: I'm Rosa.

Dev: And I'm Dev, here with Taro.

Rosa: We're looking at the paper: "Over-the-Air Consensus-based Formation Control of Heterogeneous Agents".

Dev: Here's the gist of the paper. Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees investigates the formation control problem of heterogeneous, autonomous agents that communicate over a wireless multiple access channel.

Rosa: The paper derives a communication-rate based sufficient condition for convergence to a prescribed formation and provides a geometry-aware refinement showing how favorable tracking transients can relax the required condition.

Dev: Problem Formulation and Assumptions The study deals with heterogeneous autonomous agents moving on a two-dimensional plane, where agent dynamics are described by equations involving position and control input.

Rosa: A key assumption is that each agent tracks constant reference positions exponentially, which accommodates a broad class of platforms such as unicycle and differential drive models.

Paper summary: Dev: So, looking at the whole paper, "Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees," the authors essentially propose a consensus strategy for heterogeneous agents that only needs them to track constant reference positions exponentially.

Rosa: That’s right. The main contribution is providing a communication rate based sufficient condition for convergence, but they also provide this geometry-aware refinement that lets you use tracking control to relax the original requirement on how often you have to communicate >

Taro: It seems like the paper gives us a practical way to manage the trade-off between communication frequency and geometric accuracy, which is something we see a lot in real-world autonomous systems where things aren't perfectly modeled >

Dev: The implication for someone listening outside the lab is that they can design a system where you don't have to rely solely on perfect channel knowledge or constant high-frequency updates if you use the right tracking control laws >

Rosa: It moves the focus away from just brute-force communication overhead and toward leveraging how well your agent tracks its target, which is a much more realistic constraint for physical hardware >

Taro: I think it’s about showing that even with imperfect wireless communication, you can achieve the desired formation as long as the transient geometry is managed correctly >

Dev: That’s the main idea of this work on Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees >

Conclusion: Rosa: So, we're wrapping up this look at "Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees." Basically, they figured out a way for different robots to coordinate their positions using just radio signals from each other.

Dev: Right. They used this wireless setup, what we call WMAC, where agents can send messages at the same time and mix them together to figure out where everyone should be without needing constant direct contact between every pair.

Taro: The big deal here is that they provide two different ways to prove things work. First, a condition based purely on how often the agents talk to each other, which is a communication rate requirement.

Rosa: And then they have this second part that's really interesting—it’s geometry-aware. It shows that if you use smart tracking control on your agent, you can relax that strict communication rate requirement a bit.

Dev: That means the system doesn't have to communicate as often as it used to if your agents are good at following their targets and managing those tricky movements between updates.

Taro: It’s about making the system more robust when things go wrong, like when the channel isn't perfect or when you have a bunch of different robot models trying to line up.

Rosa: The authors show that this approach can actually save a lot of effort in terms of how many antennas you need compared to older methods where every single robot had to talk directly to every other robot.

Dev: Exactly. It's about using the air smarter, not just sending more signals, which is crucial for real-world deployment where you don't want massive hardware setups just because of communication needs.

Taro: It shows that we can achieve a complex formation simply by managing the timing and the tracking errors of our individual robots well enough.

Rosa: So, it really boils down to: how much do I have to talk, and how good is my robot at keeping up with its target?

Michael Epp, Fabio Molinari, Jorg Raisch

Technische Universitat Berlin · Nimbus AI Technologies Inc. · Science of Intelligence Research Cluster of Excellence

eess.SY, cs.SY

Submitted: 2026-03-10

Updated: 2026-11-08

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 80/100

The gist: Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees investigates the formation control problem of heterogeneous…

Key concepts

Heterogeneous Agents
These are autonomous agents that have different dynamics or types, such as unicycle or differential drive models. The control scheme is designed to ensure these diverse agents can successfully maintain a desired relative formation despite their individual differences.
WMAC Model
This model describes the wireless communication medium, allowing multiple agents to broadcast simultaneously. It uses the superposition property of the channel, meaning received signals are a weighted sum of all simultaneous transmissions, which is key for the consensus computation.
Communication-Rate Based Condition
This is a theoretical requirement based on how often agents communicate. The paper finds a minimum required communication interval to guarantee convergence; however, this condition can be made more flexible by incorporating geometric information about agent movement.

Terminology

Summary

Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees investigates the formation control problem of heterogeneous, autonomous agents that communicate over a wireless multiple access channel. p. 0

Where the paper says it

1 Over-the-Air Consensus-based Formation Control of Heterogeneous Agents: Communication-Rate and Geometry-Aware Convergence Guarantees Michael Epp, Fabio Molinari, Jorg Raisch ¨ Abstract—This paper investigates the formation control problem of heterogeneous, autonomous agents that communicate over a wireless multiple access channel. Instead of avoiding interference through orthogonal node-to-node transmissions, we exploit the superposition property of the wireless channel to compute, at each receiver, normalized convex combinations of simultaneously broadcast neighbor signals.

The paper derives a communication-rate based sufficient condition for convergence to a prescribed formation and provides a geometry-aware refinement showing how favorable tracking transients can relax the required condition. p. 0
Where the paper says it

We derive a communication-rate based sufficient condition that guarantees convergence to a prescribed formation. We then provide a geometry-aware refinement showing how favorable tracking transients can relax the required condition.

Problem Formulation and Assumptions

The study deals with heterogeneous autonomous agents moving on a two-dimensional plane, where agent dynamics are described by equations involving position and control input. A key assumption is that each agent tracks constant reference positions exponentially, which accommodates a broad class of platforms such as unicycle and differential drive models. p. 0

Where the paper says it

At every communication instant, agents update their reference positions from these aggregates, and track the references in continuous time between updates. The only assumption on the agent dynamics is that each agent tracks constant reference positions exponentially, which accommodates a broad class of platforms.

The agents exchange information at sampled times with uniformly bounded inter-sampling times, meaning there exist T, ˇ Tˆ ∈ R>0, such that ∀k ∈ N0, Tˇ ≤ tk+1 − tk ≤ T. p. 1
Where the paper says it

The agents exchange information at sampling times tk ∈ R≥0, k ∈ N0, with uniformly bounded inter-sampling times, i. , there exist T , ˇ Tˆ ∈ R>0, such that ∀k ∈ N0, Tˇ ≤ tk+1 − tk ≤ T .

The objective is to achieve a prescribed formation described by displacement vectors di in the sense that limt→∞ pi(t) = c(t) + di, where c(t) is the centroid of the formation. p. 1
Where the paper says it

ˆ We aim to achieve a prescribed formation described by displacement vectors di ∈ R 2 . This is achieved asymptotically if, ∀i ∈ N , limt→∞ pi(t) = c(t) + di , (3) where c(t) ∈ R 2 is the so-called centroid of the formation.

Communication Model and Channel Properties

The study adopts the WMAC model to describe the superposition property of the wireless medium, allowing agents to broadcast simultaneously and compute normalized convex combinations from simultaneous transmissions. The received superimposed value at a receiving agent is defined as yi(tk) = Xj∈Ni(tk) ξij (tk)αj (tk), where ξij (tk) are unknown time-varying channel coefficients. p. 1

Where the paper says it

The obtained superimposed value at the receiving agent is yi(tk) = X j∈Ni(tk) ξij (tk)αj (tk), where ξij (tk) ∈ (0, 1] are unknown time-varying channel coefficients. Since the received superimposed value is a result of power modulation, the channel coefficients are assumed to be positive.

A critical result is that the matrix Hk, with [Hk]ij = hij,k, is row-stochastic. p. 2
Where the paper says it

Corollary 1. For all k ∈ N0, the matrix Hk, with [Hk]ij = hij,k, is row-stochastic.

Assumption 2 imposes a standard connectivity requirement for consensus over time-varying networks by stating there exist L ∈ N and µ ∈ R > 0 such that for all k in N0 τ1(Hk+L−1 · · · Hk) ≤ 1 − µ. p. 2
Where the paper says it

There exist L ∈ N and µ ∈ R > 0 such that for all k ∈ N0 τ1(Hk+L−1 · · · Hk) ≤ 1 − µ. Assumption 2 is a standard connectivity requirement for consensus over time-varying networks.

Convergence Analysis and Conditions

Theorem 1 establishes a general convergence condition based on the time between communication instants, requiring T >ˇ Tˇ∗:= -1/λˇ ln (1 + µ) / (L − 1) 2Cˆ. This bound can be conservative in practice, leading to the development of a geometry-aware refinement. p. 0

Where the paper says it

Rather than assuming identical agent dynamics, we require only that each agent’s position is exponentially stabilizable to any constant reference, covering a broad range of platforms, including unicycle and differential drive models [15] and car-like models [16]. For the resulting jumpflow system, we derive a communication-rate based sufficient condition for convergence to the desired formation and provide a geometry-aware refinement that captures how the tracking transient can relax the required communication-rate bound, thereby generalizing earlier formation control results in [12] and [14].

Theorem 2 provides a sufficient condition that relaxes Assumption 1 by introducing tracking control, requiring σi,kβi < 1/2 (1 + ˇσLµ) / (L − 1) for all i ∈ N and k ∈ N0. p. 5
Where the paper says it

In this sense, beyond enlarging Tˇ, the tracking controller itself provides a second lever for guaranteeing consensus and thereby convergence to the desired formation. Consequently, Theorem 2 generalizes not only Theorem 1, but also the findings in [12], where the tracking control satisfied ∀k, ∀i, σi,kβi = 0 for any choice of σi,k ∈ (0, 1], i.

This condition allows the system to achieve the desired formation even when increasing the inter-communication time is undesirable. p. 3
Where the paper says it

The jump-flow system in (4) to (6) therefore converges if the time in between communication instants is sufficiently large. However, the requirement in (13) can be conservative in practice since increasing the inter-communication time may be undesirable, as in many settings the sampling schedule is fixed, and increasing Tˇ might lead to reduced responsiveness of the system.

Geometric Interpretation and Simulation

Theorem 2 offers a geometric interpretation where the product βiσi,k acts as a radius multiplier quantifying how far an agent may deviate from the segment point si,k+. The parameter σi,k can be interpreted as a certificate of the tracking control on flow interval k. p. 5

Where the paper says it

(20) Thus, the product βiσi,k acts as a radius multiplier that quantifies how far the agent may deviate from the segment point si,k+ . The parameter σi,k can be interpreted as a certificate of the tracking control on flow interval k.

Simulations show that increasing µ introduces a stronger tangential component causing trajectories to deviate further from the line segment between the previous position and the updated reference, which makes it harder to satisfy the inequality in Theorem 2. p. 6
Where the paper says it

This behavior is consistent with the geometric interpretation of Theorem 2. Increasing µ introduces a stronger tangential component, causing trajectories to deviate further from the line segment between the previous position and the updated reference, which in turn makes it harder to satisfy the inequality in Theorem 2.

Convergence is recovered by increasing the time in between communication instants, matching prediction of Theorem 1. p. 6
Where the paper says it

However, convergence is recovered in the third experiment by increasing the time in between communication instants (µ = π 2T , T = 1 s). This matches the prediction of Theorem 1, which guarantees convergence for sufficiently long flow intervals.

The protocol also provides a substantial reduction in the number of required orthogonal channels compared to traditional interference-avoiding node-to-node communication protocols. p. 6
Where the paper says it

The protocol exploits the superposition property of the wireless channel to compute normalized convex combinations from simultaneous transmissions. This yields a substantial reduction in the required number of orthogonal channels when compared to traditional interference-avoiding node-to-node communication schemes.

Conclusion and Future Work

The paper proposes a consensus-based formation control scheme for heterogeneous agents by only requiring exponential stabilizability of the position with respect to a constant reference. Convergence can be ensured by choosing the time in between communication updates sufficiently large or through an appropriate choice of tracking control laws that yield favorable transient geometry. p. 6

Where the paper says it

In particular, convergence can be ensured by choosing the time in between communication updates sufficiently large. Moreover, convergence can also be established through an appropriate choice of tracking control laws that yield favorable transient geometry.

Future work plans to incorporate additional adverse communication effects, including additive noise and halfduplex communication. p. 6
Where the paper says it

This yields a substantial reduction in the required number of orthogonal channels when compared to traditional interference-avoiding node-to-node communication schemes. In future work, we plan to incorporate additional adverse communication effects, including additive noise and halfduplex communication.

The OTA broadcast protocol exploits the superposition property of the wireless channel to compute normalized convex combinations from simultaneous transmissions. p. 6
Where the paper says it

A key element of the approach is the OTA broadcast protocol. The protocol exploits the superposition property of the wireless channel to compute normalized convex combinations from simultaneous transmissions.

The employed protocol provides a substantial reduction in the number of required orthogonal channels when compared to traditional interference-avoiding node-to-node communication schemes. p. 6
Where the paper says it

The protocol exploits the superposition property of the wireless channel to compute normalized convex combinations from simultaneous transmissions. This yields a substantial reduction in the required number of orthogonal channels when compared to traditional interference-avoiding node-to-node communication schemes.

REFERENCES

Bares 8

Molinari 17

Raisch 17

Epp 3

Molinari 4172

Epp 4

Molinari 4172

Raisch 8

Molinari 4172

Epp, Molinari and Raisch are with the Con- ¨trol Systems Group, Technische Universitat Berlin, Germany. Email: michael. p. 0

Where the paper says it

In the latter case, distributed consensus mechanisms are a Michael Epp, Fabio Molinari and Jorg Raisch are with the Con- ¨ trol Systems Group, Technische Universitat Berlin, Germany. Email: ¨ michael.

epp@tu-berlin. p. 0
Where the paper says it

On the other hand, in centroid-based formation control, agents agree on a common reference point and realize the formation through prescribed displacement vectors with respect to that reference [6], [7]. In the latter case, distributed consensus mechanisms are a Michael Epp, Fabio Molinari and Jorg Raisch are with the Con- ¨ trol Systems Group, Technische Universitat Berlin, Germany.

de Fabio Molinari is also with Nimbus AI Technologies Inc. p. 0
Where the paper says it

epp@tu-berlin. de Fabio Molinari is also with Nimbus AI Technologies Inc.

, Seattle, USA. p. 0
Where the paper says it

de Fabio Molinari is also with Nimbus AI Technologies Inc. , Seattle, USA.

Jorg Raisch is also with Science of Intelligence (SCIoI), ¨Research Cluster of Excellence, Berlin, Germany. p. 0
Where the paper says it

, Seattle, USA. Jorg Raisch is also with Science of Intelligence (SCIoI), ¨ Research Cluster of Excellence, Berlin, Germany.

Email: raisch@control. p. 0
Where the paper says it

Jorg Raisch is also with Science of Intelligence (SCIoI), ¨ Research Cluster of Excellence, Berlin, Germany. Email: raisch@control.

tu-berlin.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Over-the-Air Consensus-based Formation Control for Heterogeneous Agents: The improved system can achieve convergence to a prescribed formation using only sampled wireless communication, leveraging the superposition property of wireless channels instead of requiring orthogonal node-to-node transmissions.

  2. Communication Rate Based Convergence Guarantee: The system can guarantee convergence to a desired formation by deriving a communication-rate based sufficient condition that accounts for the time in between communication instants being sufficiently large, or by incorporating favorable tracking transients.

  3. Geometry Aware Tracking Control: The system can be designed with tracking control laws that yield favorable transient geometry, allowing convergence even when the required communication-rate bound is relaxed through geometric properties of the tracking behavior.

Abstract

This paper investigates the formation control problem of heterogeneous, autonomous agents that communicate over a wireless multiple access channel. Instead of avoiding interference through orthogonal node-to-node transmissions, we exploit the superposition property of the wireless channel to compute, at each receiver, normalized convex combinations of simultaneously broadcast neighbor signals. At every communication instant, agents update their reference positions from these aggregates, and track the references in continuous time between updates. The only assumption on the agent dynamics is that each agent tracks constant reference positions exponentially, which accommodates a broad class of platforms. Under this assumption, we analyze the resulting jump-flow system under time-varying communication graphs and unknown channel coefficients. We derive a communication-rate based sufficient condition that guarantees convergence to a prescribed formation. We then provide a geometry-aware refinement showing how favorable tracking transients can relax the required condition. Simulations with unicycle agents illustrate the theoretical results and demonstrate a substantial reduction in the number of required orthogonal transmissions compared to interference-avoiding node-to-node communication protocols.

Related papers