ISCC test results: communication and ranging
Results reviewed September 22, 2026. Hardware tests: September 20, 2026. Scope: indoor LD-50 laboratory prototype.
Published ISAC report: Read the technical report on Zenodo · DOI: 10.5281/zenodo.22904657
Milkyway Center is developing a system that uses one received signal for both communication and sensing. The longer-term aim is to connect those observations to control decisions. We call that goal ISCC: Integrated Sensing, Communication, and Control.
ISAC demonstrated in real hardware tests. The LD-50 bench measurements confirm integrated sensing and communication under the reported indoor conditions: the receiver decoded information and estimated range from the same incoming optical stream. This is a measured result within the broader ISCC project.
The latest laboratory work demonstrates a practical step toward that goal: the receiver recovered information while estimating distance from the same incoming frame stream. In a run lasting about ten minutes, it processed 150,000 frames without losing a frame, delivered 149,860 accepted frames, and recorded no errors in the decoded information it delivered.
Those results include error correction and a receiver confidence check. The uncorrected signal still contained errors. The distinction matters: this is a measured prototype result under specific conditions, not a promise of an error-free channel in every environment.
What the latest test measured
The September 20 continuous-stream test used the indoor bench setup, aimed at a fixed direction. The transmitter repeatedly played a table of 1,024 codewords, with random information encoded using Golay(24,12) error correction. The receiver processed each arriving frame, checked whether it should be accepted, and estimated range from reference samples carried in that same stream.
| Measurement | September 20 result | What it means |
|---|---|---|
| Frames processed | 150,000 | A run lasting 605.24 seconds, about ten minutes |
| Lost or malformed frames | 0 | The recorded frame sequence remained intact during this run |
| Accepted delivery | 149,860 / 150,000, or 99.907% | 140 frames were not accepted for delivery |
| Decoded frame errors | 0 observed | Error correction recovered the tested information; delivered frames also had zero observed errors |
| Raw bit error rate | 0.000637, or 0.0637% | Errors were present before error correction |
| Earlier receiver on the same frames | 0.00587, or 0.587% | A comparison using the same received observations |
| Delivered information rate | 2.971 kbit/s | Correctly delivered information divided by total elapsed time |
| Range from the same stream | Median 93.3 inches, about 2.37 m | Based on 2,047 accepted range estimates |
| Range spread | Median absolute deviation 0.85 inches, about 2.16 cm | A measure of variation, not an absolute accuracy claim |
The important result is the combination: communication and range were available from the same stream. Distance was not inferred from the decoded message. At the same time, not every frame produced an accepted distance estimate, and the reference portion used for range did not carry information bits.
How this compares with the earlier run
An earlier September 20 trial processed 30,000 frames over 121.1 seconds, also without lost or malformed frames. It accepted 29,959 frames, or 99.863%, with zero observed decoded errors and zero wrong frames delivered. Correct information throughput was 2.970 kbit/s.
That earlier trial produced 337 range estimates at 2.78 estimates per second, with a median of 93.4 inches and a single-estimate median absolute deviation of 1.08 inches. Its raw bit error rate was 0.0007903, compared with 0.006315 for the earlier receiver on the same frames.
The later, longer run supports the same basic finding. It does not turn either observation into an unlimited reliability guarantee. Each result remains attached to its own test duration, payload, receiver configuration, and laboratory conditions.
From decoded bits to a readable message
The same day's integration work also tested a short message. The receiver filled all 37 message slots, reconstructing 444 information bits that matched the transmitted message bit for bit. The recorded delivered-error count was zero, with a reported information rate of 2.964 kbit/s and a same-stream range of 93.2 inches.
The interface now distinguishes a received slot from a missing or rejected one. In subsequent repeated-message checks, the ledger records 8,262 repeat comparisons at the aimed direction and 21,932 comparisons cumulatively after other directions were included, without a disagreement. Repeated playback filled all message slots in the tested directions.
This establishes message reconstruction and visible loss handling for those trials. It does not establish recovery from every interruption. Stronger loss conditions were not captured in that particular check, and agreement between repetitions is not by itself proof of absolute correctness.
What remains to be tested
Fresh traffic and a second device. The continuous tests used a repeating codeword table. Updating the message stopped, rewrote, and restarted playback; uninterrupted replacement with fresh payload data was not validated. These measurements used the laboratory's returned optical signal. They do not demonstrate a separately transmitting and receiving pair of independent devices.
Range accuracy and difficult scenes. The reported range medians and spreads describe the tested data. A repeatable estimate can still contain a scale error. The ledger identifies approximately 2.8% unresolved absolute scale offset at this operating point. Instrument-related false lines also affected some short intervals, especially when only a few estimates were available. Wider claims need independent distance references, sufficient accepted observations, and checks in more difficult scenes.
Outdoor operation. The September 21 audit treats 6 m as the boundary of existing relevant distance evidence, not a physical limit of the instrument. Longer distances, moving targets with independently known speed, sunlight, temperature changes, and road scenes still require their own physical validation. Calculated range budgets and simulated targets are not outdoor test results.
Complete system validation. The September 21 software work introduced a new acquisition and observation path and passed 11 offline tests using simulated hardware. It did not complete a new hardware validation of the full processing method, timing verification, feedback loop, and sensing chain. Earlier measurements retain their original methods and conditions; they are not relabeled as passing newer requirements.
Separate design and simulation results
The September 16 digital-capture work passed five scenarios using simulated clocks and diagnostic input. These checked complete-frame delivery, missing data, buffer overflow, and recovery. No physical board was programmed, and final implementation timing was not closed. This is functional design verification.
A separate control simulation completed all 50 agents' goals in each of seven scenarios. Its calculated continuous minimum separation was 5.098806 m. The model used ideal nearby-position observations and registered parallel or circular paths. It did not include real aircraft, wind, tracking error, or a complete optical sensing chain; the result applies to those modeled conditions.
The higher-rate optical project retains a 1 Tbps design target. Its timing audit provides a conditional capacity budget, while optical simulations still report failures under phase disturbance and incomplete integration. These records do not establish a manufactured chip, a working 1 Tbps link, or field deployment.
Where control and security fit
The September 20 hardware tests concern communication, range, and receiver integration. They do not demonstrate autonomous vehicle control, safe obstacle avoidance, material identification, or a completed network-scale control system.
Earlier tensor-field coordination results were simulations. Earlier security statements concerned a separate protocol model and its assumptions. Neither should be read as a new capability established by this hardware run. Future updates will identify which results come from physical measurements, recorded-data analysis, simulation, or proposed work.
For research discussions, demonstrations, or questions about the test conditions, contact yuanjimaster@gmail.com. A public evidence summary lists the test records and their checksums so these results can be tied to specific archived artifacts.
Read the September 2026 evidence summary for test scope, artifact names and SHA-256 checksums.
Published report and citation
Huo, F. (2026). ISAC Demonstrated in Real Optical Hardware: LD-50 Bench Results from September 20, 2026 (Version 1.0) [Technical report]. Zenodo. https://doi.org/10.5281/zenodo.22904657
The archived report includes the real hardware measurements, aggregate results, source fingerprints, and test limitations. It covers the ISAC bench demonstration; the separate control simulations and future design targets on this page are outside that report's scope.