SemOCC: Task-Oriented Optical Camera Communication for Robust Robotic Command Signaling
Year: 2026
Type: article
Abstract: Radio Frequency (RF) networks drive Industry 4.0 robotic fleets, but they suffer from electromagnetic interference and congestion. Optical Camera Communication (OCC) provides an interference-free complement. However, conventional OCC recovers bit streams through temporal intensity modulation. Its performance degrades sharply when motion blur, occlusion, and low signal-to-noise ratio (SNR) occur simultaneously. The link also forces a reliability-latency trade-off through retransmissions. Existing OCC work optimizes bit-level fidelity rather than task-level outcomes. This leaves a gap for command-oriented robotic signaling. This paper proposes Semantic Optical Camera Communication (SemOCC), a semantic-inspired task-oriented signaling framework. The transmitter encodes each complete robotic command as a structured light pattern using a Digital Micromirror Device (DMD). The receiver uses a lightweight Convolutional Neural Network (CNN) to infer the command directly from the captured image, without recovering individual bits. We measure performance through task-oriented mutual information and semantic efficiency, both computed from empirical confusion matrices. We compare SemOCC against three baselines under a unified channel model: a conventional bit-centric OCC link with on-off keying, rolling-shutter capture, and Hamming coding; a Hadamard normalized cross-correlation scheme; and dynamic fiducial markers. Under severe combined impairments, SemOCC achieves an 89.3% task-success rate at a deterministic 46 ms latency. The bit-centric OCC link reaches only 21.5% in single-frame mode, or 38.4% with retransmission at a 138 ms 95th percentile latency. These simulation results suggest that command-level spatial encoding and direct task inference can improve the reliability-latency operating point of OCC links for industrial robotic command delivery.
Source: IEEE Access
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W7204122244