Chicane review

1. Choose a recorded day or upload UART files

No data loaded yet.

Files are processed in this browser. Saved detections can also load their complete UART capture from Firebase.

2. Detections and light sequences

Load a day or UART files to begin.

Review references B1, B2… number detections from oldest to newest within the selected day, before filtering or sorting. They refer to the whole detection. Clear the search to show all rows again.

Direction filters select detections containing a report in that direction; mixed-direction detections keep all their raw reports visible. Chicane strips: lights remain off until an ID qualifies on its second consecutive report (at least one report ≥8 km/h); a missing ID or polling gap over 0.6 seconds resets confirmation. Reports may confirm outside 100 m, but cannot activate lights there. Earlier reports are never retrospectively sent to the lights. Then: nearest toward target within 100 m; a new target at 10–30 mph starts green immediately, above 30 starts amber. Amber changes to green after two consecutive reports at or below 30. Once green, the overlapping group stays cleared even if it accelerates above 30 mph. Green continues while a confirmed group member moves at least 8 km/h (about 5 mph), including away movement inside 100 m. When the group slows or disappears, it gets two more seconds of green; confirmed movement before expiry cancels this grace. Completed cars are then ignored, and a distinct new arrival can start its own amber/green sequence while an old stopped car remains. Nearby replacement tracking IDs are associated by position and speed; radar IDs cannot guarantee physical car identity. Two seconds without in-zone reports clear the old group history. Time axes include the final predicted grace. Away reports are dashed at 50% opacity; they cannot start a clearance but can keep an already-cleared group green. Toward reports are solid. The red line marks 30 mph. Saved samples stay unchanged on the left. Adding full recordings shows the complete input and a third graph of live accepted reports. The third graph contains only fresh measurements after two-report confirmation, toward and within 100 m, meeting the 10 mph entry rule (lower speeds remain visible for the active target). Confirmation samples are never backfilled. Its light strip comes from the same simulation run, not a second filtering pass. This reproduces the configured model, not measured hardware latency or a verified Pi deployment. Both columns share scales; matching measurements align their display when available; this alignment does not affect light decisions. Saved samples omit reports and are an incomplete replay; use the full recording to assess live decisions. Filters and B references continue to use the saved detections.