Constrained Bandwidth Observation Sharing for Multi-Robot Navigation in Dynamic Environments via Intelligent Knapsack

TL;DR

Confronts bandwidth-limited multi-robot navigation in dynamic environments by introducing Intelligent Knapsack (iKnap), which models pairwise observation sharing as graph-based multi-agent inference and solves a 0/1 knapsack to maximize information value under a bandwidth budget $B$. The approach is implemented atop a centralized infrastructure with a pairwise communication scheme, using a KL-divergence-based belief update and a trajectory-aware distance metric to form the utility $\theta$, and is solved in real time via dynamic programming with complexity $O(n^2B)$. Extensive ROS2/RMF simulations in a warehouse show that iKnap yields navigation performance gains over broadcast-based and learning-based baselines, while using bandwidth more efficiently, especially under low-resource and high-uncertainty conditions. The work demonstrates practical feasibility for robust multi-robot collaboration in real-world navigation tasks and lays groundwork for generalized, learning-augmented utility design and factory-scale deployments.

Abstract

Multi-robot navigation is increasingly crucial in various domains, including disaster response, autonomous vehicles, and warehouse and manufacturing automation. Robot teams often must operate in highly dynamic environments and under strict bandwidth constraints imposed by communication infrastructure, rendering effective observation sharing within the system a challenging problem. This paper presents a novel optimal communication scheme, Intelligent Knapsack (iKnap), for multi-robot navigation in dynamic environments under bandwidth constraints. We model multi-robot communication as belief propagation in a graph of inferential agents. We then formulate the combinatorial optimization for observation sharing as a 0/1 knapsack problem, where each potential pairwise communication between robots is assigned a decision-making utility to be weighed against its bandwidth cost, and the system has some cumulative bandwidth limit. We evaluate our approach in a simulated robotic warehouse with human workers using ROS2 and the Open Robotics Middleware Framework. Compared to state-of-the-art broadcast-based optimal communication schemes, iKnap yields significant improvements in navigation performance with respect to scenario complexity while maintaining a similar runtime. Furthermore, iKnap utilizes allocated bandwidth and observational resources more efficiently than existing approaches, especially in very low-resource and high-uncertainty settings. Based on these results, we claim that the proposed method enables more robust collaboration for multi-robot teams in real-world navigation problems.

Figures (6)

• Figure 1: As robots $\mathcal{A}_1$ and $\mathcal{A}_2$ navigate through a warehouse environment, $\mathcal{A}_1$ observes moving obstacle $\mathcal{S}_1$ and shares its observation with $\mathcal{A}_2$ over the air. Consequently, $\mathcal{A}_2$ is made aware of a previously invisible potential collision with $\mathcal{S}_1$ and can plan accordingly.
• Figure 2: Example scenario involving a robot team $A_1, A_2, A_3, A_4$ and a dynamic obstacle subject $S_1$ only visible by $A_1$ due to a wall obstruction. There is an impending collision between $A_3$ and $S_1$. In (a), all agents routinely upload their states and beliefs to the infrastructure $\mathcal{I}$, and $A_1$ additionally uploads its observation of $S_1$ to $\mathcal{I}$. Following this process, the proposed pairwise communication scheme is presented in (b), where $A_1$'s observation is shared only with $A_3$ to enable safe re-planning around $S_1$ while maintaining efficient communication. Alternatively, (c) depicts the traditional broadcast-based communication scheme, where $A_1$'s observation is shared with the entire robot team, despite the information's irrelevance to the decision-making of $A_2$ and $A_3$.
• Figure 3: Example warehouse environment generated with the Open-RMF Traffic Editor tool.
• Figure 4: Capability studies comparing makespan (s) for iKnap (blue), OCBC (green), random pairwise (red), random broadcast (orange), and no communication (purple) baselines. We measure performance vs. (a) robot/obstacle count $n+m$, (b) sensor noise level, (c) communication latency, and (d) environment complexity (number of walls). Each configuration's measurements are averaged across multiple randomized trials.
• Figure 5: Optimizer runtime (s) vs. agent/obstacle count for iKnap (blue) and OCBC (green).