Networked Iterative Learning Control Using a Novel Bit-Efficient Shell Quantization Method
Communication constraints in networked control systems (NCSs) severely limit the transmission of plant outputs, controller commands, and learning variables. Conventional finite-level quantizers distribute their quantization cells over the entire admissible region, which becomes inefficient when the transmitted signal is vector-valued and evolves dynamically. This article proposes a finite-level shell quantization method and applies it to networked iterative learning control (ILC) as a representative scenario. The proposed shell quantization comprises a shell quantizer, two preprocessing and reconstruction mechanisms, and an encoding-decoding pair. The shell quantizer places all nonzero reconstruction points on a single spherical shell, allocating finite quantization levels primarily to directional information; radial information is represented by a transmitted radial-layer index and recovered at the decoder. Two radial recovery mechanisms are developed: the additive mechanism uses uniformly spaced radial layers, and the multiplicative mechanism uses logarithmically spaced radial layers. For networked ILC, the proposed quantization is incorporated into an encoding-decoding structure, where the shell-quantized point and the radial-layer index are jointly transmitted for output reconstruction. Convergence analysis reveals that the additive mechanism yields bounded tracking-error convergence with an explicit error bound, whereas the multiplicative mechanism leads to region-dependent cyclic boundedness, making it suitable for signals with large amplitude variations. Simulations on a robot joint model validate the analysis and demonstrate that, under the same quantization-bit budget, the proposed shell quantization achieves improved tracking performance over conventional quantizers.
Authors
- Dong Shen (ORCID: https://orcid.org/0000-0003-1063-1351)
- Daniel W. C. Ho (ORCID: https://orcid.org/0000-0001-9799-3712)
- Taojun Liu (ORCID: https://orcid.org/0009-0002-6560-7797)
Institutions
- City University of Hong Kong (HK)
- Renmin University of China (CN)
Publication Details
- Journal
- IEEE Transactions on Cybernetics
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1109/tcyb.2026.3750031
- Primary Topic
- Iterative Learning Control Systems
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China