Back to the portfolio

AI-learned encoder-decoder for MIMO wireless base stations

Communications & RFComputing, Software & AIElectronics & SemiconductorsLab validated

This technology uses machine learning to improve how data is sent and received over wireless networks that use multiple antennas simultaneously — a setup called MIMO, which is the backbone of modern 4G and 5G systems. Instead of relying on hand-engineered signal processing rules, the system trains neural networks to act as both the encoder (transmitter side) and decoder (receiver side), learning the best way to pack and unpack information through a real-world wireless channel. A key feature is that the receiver uses its reconstructed data to estimate the current state of the wireless channel — things like interference, user positions, and signal quality — and feeds that back to guide future transmissions. The whole system is trained end-to-end, meaning the transmitter and receiver jointly optimize together, rather than being designed separately as in conventional radio systems.

What you could build

A software module or chipset IP block for 5G and next-generation base stations that replaces legacy MIMO signal processing stacks with a learned encoder-decoder, improving throughput and reliability in dense urban or high-interference environments; target buyers are telecom equipment manufacturers like Ericsson, Nokia, and Samsung Networks, as well as O-RAN software vendors.

Who in Virginia should care

Northern Virginia's dense concentration of data center operators, federal telecom contractors, and DoD communications programs creates demand for advanced wireless processing IP, particularly for private 5G and resilient military communications.

Readiness: Lab validated

Concept — described but not yet demonstrated. Lab validated — supported by experimental results in the patent. Prototype likely — the text describes a built, working embodiment.

Readiness is inferred from the patent text, not from a lab visit.

The record

Inventors
Timothy James O`Shea, Tugba Erpek
Granted
July 5, 2022
Status
Granted patent
Patent number
11381286

Ready to talk?

Virginia Tech Intellectual Properties handles licensing for this technology.

VTIP contact coming shortly

Prosim summaries are generated from public patent text and are not legal advice.