Lightweight Change Detection for Adaptive Collective Perception in Dynamic Environments

Thiemen Siemensma; Tiago Travassos; Jiacheng Zhang; Bahar Haghighat

cs.ROOct 25, 2026

Lightweight Change Detection for Adaptive Collective Perception in Dynamic Environments

Authors: Thiemen Siemensma, Tiago Travassos, Jiacheng Zhang, Bahar Haghighat

Organizations: Faculty of Science and Engineering, University of Groningen, The Netherlands

Abstract

Swarm robotics enables scalable inspection through collective perception, where robots infer environmental properties from distributed observations. A common benchmark considers robots exploring black-white tiled environments and determining the majority tile type. While most work studies static environments, dynamic environments remain underexplored. We study collective perception in dynamic environments where the tile proportion and the spatial correlation of surface patterns may change over time. To enable implementation on resource-constrained robots, we integrate a lightweight change detection mechanism into a Bayesian collective decision-making framework that resets robot beliefs when distribution shifts are detected. The proposed detector requires constant memory and constant per-sample computation, unlike two recent competitive approaches that store the full sample stream, making it suitable for memory-constrained robots. We further introduce a method that detects spatially correlated sampling and adapts sampling intervals online to mitigate correlation effects. Our simulation results in Webots show competitive change detection delays, more consistent precision across communication settings, and improved decision accuracy under changing spatial correlations.

Figures & tables

Panel of Figure 1 — Figure 1: We consider three possible changes in our study of dynamic environments: (a) fill ratio changes from predominantly black to predominantly white ( $f_{0}=0.3\rightarrow f_{1}=0.7$ ). (b) Spatial correlation changes due to clustered tiles. (c) Spatial correlation changes due to feature size changes.

Panel of Figure 2 — Figure 2: Simulations use the vibration-sensing robot (a) within a bounded environment (b). Scenarios (c): fill ratio change detection, spatial-correlation detection, and adaptation to improve accuracy.

Panel of Figure 3 — Figure 3: Belief trajectories (100 runs) for CUSUM ( $\theta_{p}=0.6$ ), DBB, and DBBCPD over 1200 s with a fill ratio inversion at $t=600$ s, for all-to-all and reduced (3 body-length range) communication. In easier environments (larger $\Delta f$ ), CUSUM shows improved belief values over time. In harder environments (smaller $\Delta f$ ), CUSUM achieves superior beliefs only with increased communication.

Range	$\Delta f$	Precision			Recall			Delay ( $\mu\pm\sigma$ ) [s]
Range	$\Delta f$	CUSUM	DBB	DBBCPD	CUSUM	DBB	DBBCPD	CUSUM	DBB	DBBCPD
All-to-all	$0.4$	0.65	0.61	0.20	1.00	1.00	1.00	$20\pm 6$	$93\pm 21$	$21\pm 9$
	$0.2$	0.55	0.51	0.20	0.99	0.99	0.96	$49\pm 26$	$197\pm 65$	$50\pm 68$
	$0.1$	0.52	0.33	0.13	0.88	0.80	0.98	$131\pm 97$	$269\pm 142$	$93\pm 108$
Reduced	$0.4$	0.56	0.44	0.97	0.98	1.00	1.00	$62\pm 47$	$170\pm 50$	$81\pm 38$
	$0.2$	0.51	0.26	0.74	0.94	0.97	0.92	$147\pm 122$	$199\pm 116$	$135\pm 99$

Table 1: Evaluation metrics for CUSUM and baseline methods under various fill ratio changes (

\Delta f

) and communication ranges all-to-all and reduced (3-body-length range).

Panel of Figure 4 — Figure 4: Accuracy ( $1-|d_{f}^{*}-p|$ ) at fixed step sizes proportional to the pre-change correlation, for CUSUM ( $\theta_{p}=0.005$ ) and the baseline methods DBB and DBBCPD, in a scenario where the fill ratio ( $f=0.56$ ) remains constant but spatial correlation changes at $t=600$ s. All methods exhibit accuracy degradation as correlation increases, likely caused by false-positive fill ratio detections.

Panel of Figure 5 — Figure 5: Average tile-size estimates over 100 runs with five robots for $\ell\in{0.05,0.1,0.2}$ m. Estimates converge near $\ell$ for $d\leq\ell$ and toward $d$ otherwise. Dashed lines indicate the mean convergence point m within 10% of steady-state.

Panel of Figure 6 — Figure 6: Accuracy of the proposed CUSUM ( $\theta_{p}=0.005$ ) for fixed and adaptive step size $d$ across three environments with varying degrees of spatial correlation, while keeping the fill ratio constant ( $f=0.56$ ). Overall, CUSUM exhibits an increasing number of false positives as spatial correlation strengthens, due to correlated sampling. Adapting the step size based on the observed correlation enables more representative sampling, thereby reducing false positives.

Swarm Robotics

Swarm robotics studies how groups of robots coordinate through local sensing, communication, and interaction to accomplish shared tasks.

See all 99 papers

CoAdapt: An LLM-based Framework for Adaptive Collaborative Perception in IIoT Robotic Swarms
Sep 15, 2026Houssam Hajj Hassan, Antonia Maria Masucci, Lynda Zitoune +1Swarm Robotics Communication-Efficient Collaborative Perception
L2S
Belief Dynamics for Detecting Behavioral Shifts in Safe Collaborative Manipulation
Apr 4, 2026Devashri Naik, Divake Kumar, Nastaran Darabi +1Robotics Human-Robot Collaboration
University of Illinois Chicago
Resilient Decentralized Ergodic Coverage for Scalable Multi-Robot Systems in Unknown Time-Varying Environments
Apr 5, 2026Maria G. Mendoza, Victoria Marie Tuck, Chinmay Maheshwari +1Multi-Robot Systems Exploration-Exploitation Tradeoff
Department of Mechanical Engineering, University of California, Berkeley, CA 94720, USA · Department of Electrical and Systems Engineering, University of Pennsylvania, Philadelphia, PA 19104, USA · Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD 21209, USA +1

Sep 15, 2026J/K move · Enter open · S save

CoAdapt: An LLM-based Framework for Adaptive Collaborative Perception in IIoT Robotic Swarms

Houssam Hajj Hassan, Antonia Maria Masucci, Lynda Zitoune, Salah-Eddine Elayoubi

L2S

Swarm Robotics Communication-Efficient Collaborative Perception Multi-Robot Coordination Cooperative Perception

Open paper AskPDF arXiv

Industrial IoT environments increasingly deploy autonomous mobile robots for tasks such as material handling, product assembly, or infrastructure inspection. In such deployments, collaborative perception enables robots to share LiDAR observations and collectively construct a richer model of their environment than an individual agent could produce alone. However, industrial environments are dynamic spaces where robot positions shift continuously, network bandwidth fluctuates, and the marginal contribution of robots to perception quality varies at runtime. Existing collaborative perception approaches are designed for static participation assumptions and cannot adapt to these dynamics without sacrificing either detection precision or communication efficiency. This paper presents CoAdapt, an adaptive collaborative perception framework for IIoT robotic swarms in which a Large Language Model (LLM) serves as a runtime fusion controller, jointly deciding which robots participate in the fusion process and which fusion algorithm to apply based on the current spatial configuration and network state. The LLM reasons over structured natural language descriptions of the scene derived from raw LiDAR point clouds, requiring no taskspecific training and generalizing to unseen swarm topologies. Evaluated on the OPV2V benchmark across 25 scenarios, our approach achieves a 38% reduction in communication cost while maintaining detection precision comparable to static baseline approaches.

Abstract

Swarm Robotics

Explore similar work

CoAdapt: An LLM-based Framework for Adaptive Collaborative Perception in IIoT Robotic Swarms

Belief Dynamics for Detecting Behavioral Shifts in Safe Collaborative Manipulation

Resilient Decentralized Ergodic Coverage for Scalable Multi-Robot Systems in Unknown Time-Varying Environments