AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

RATE-DISTORTION OPTIMIZED COMMUNICATION FOR COLLABORATIVE PERCEPTION

Genjia Liu, Anning Hu, Yue Hu, Wenjun Zhang, Siheng Chen

ICLR 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Collaborative multi-agent perception

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use task entropy coding (Huffman weighted by confidence frequency) instead of fixed-length coding (ablation)RDcomm · OPV2V detection / segmentationcommunication volume: 83/57% saved (detection/segmentation)Improved
use task entropy coding instead of occurrence-frequency entropy coding (ablation)RDcomm · OPV2V detection / segmentationcommunication volume: 30/25% savedImproved
use mutual-information-driven selection instead of confidence-based or LiDAR-coverage-based selection (ablation)RDcomm · OPV2V detection / segmentationcommunication volume: 60/50% reduction (detection/segmentation)Improved
add UNet message smoothing module (ablation)RDcomm · OPV2V detection / segmentationAP50 / IoU: +4% AP50, +10% IoUImproved
use RDcomm coding with codebook 128 instead of plain VQVAE-128 (ablation)RDcomm · OPV2V BEV segmentationmIoU: 0.40 -> 0.44 mIoU; 14 -> 4 bppImproved
increase VQVAE codebook size from 32 to 128 (ablation)VQVAE baseline · OPV2V BEV segmentationmIoU: 0.38 -> 0.40; 10 -> 14 bppImproved
transmit coarse abstraction (base codes) for redundancy estimationRDcomm · DAIR-V2X detectionshare of communication volume: 9%-11% of total volumeNo clear effect
use RDcomm instead of CodeFillingcollaborative perception · DAIR-V2X / OPV2V detection, OPV2V segmentationcommunication volume at matched performance: 15/13 times less (DAIR-V2X LiDAR/camera), 54/108 times less (OPV2V), 8 times less (segmentation)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy