Four HIVE TIPs in 30 seconds
A two-robot scripted planner completed four HIVE TIPs before the end of a 30-second AUTO fixture. Its cleanest witness, Red seed 31260975, finishes at 2.542, 10.200, 18.858 and 29.317 seconds. All four are completion events strictly before the cutoff. [1]
This is an alliance result with two active robots, not a solo record. The four-TIP sequence appeared in two of 48 deliberately selected seeds; none of the 12 unselected paired seeds reached four. We report both results because finding a possible sequence and demonstrating reliability are different research questions. [1][2]
Watch the four-TIP AUTO opening
The first 30 seconds of the separately recorded full-match continuation, seed 31260975. Red completes four AUTO TIPs; choose Full match to watch the continuation to 12 total TIPs and a 256-211 finish. This recording is not the independent discovery fixture.
Loading the recorded match...
[*] Starting score: Each alliance begins with 10 modeled field-content points: 4 GARDEN points (four starting POLLEN at 1 point each) plus 6 HIVE-retained points (three starting NECTAR at 2 points each). These are already present at setup, not earned by robot actions. We keep the original recorded totals; contents points change as pieces move and are not a fixed 10-point bonus. Replay evidence.
Starts paused at 1x. Positions are original recorded frames, not interpolated or resimulated. Exact TIP event timestamps can fall between frames; event jumps show the first recorded frame at or after completion.
Playback provenance
Original recording information appears when playback loads.
Download the sanitized replay dataSource [3]. Uncalibrated simulation; no claim of physical robot performance.
More useful work on the same twelve seeds
Mean Red AUTO TIPs in separate, unselected paired seeds 41270001-41270012. The scripted baseline and staged planner use the same physical model. The planner improves the mean without reaching four in this sample.
Parallel collection beats a serialized handoff
The lower-wall robot shoots its four preloads at the initial HIVE face. Its opposite-wall partner waits for the first TIP, then interleaves its own preload shots with extraction from the nearby FLOWER, freeing inventory space as it works. Meanwhile, the first robot gathers the lower-wall GARDEN POLLEN for the next volley. [1]
The second robot continues to another FLOWER supply. Both robots gather and shoot conserved loose POLLEN and released NECTAR; FLOWER scoring itself stays disabled. The fourth TIP is not explained by four fresh GARDEN balls alone: recycling and physically picked-up released pieces contribute to it. [1]
This experiment selects legal observed actions in unchanged v10 physics. It is an offline scripted controller, not an RLE-trained or promoted MAPPO actor. The model's 50% open-field AUTO pickup rejection remains enabled. [1]
Keep the selection bias next to the headline
The discovery set was selected because archived recordings had already shown three Red AUTO TIPs on those seeds. The staged controller reached four on seeds 31260975 and 31261165. The second witness completed at 29.575 seconds but required modeled teammate-separation retreats; the first had no such retreats in its 30-second fixture. [1]
On the twelve unselected paired seeds, mean Red AUTO TIPs rose from 2.083 to 2.917: ten improvements, two ties and no regressions. Both arms had zero PIN and teammate-overrun flags. This is a small, encouraging model comparison, not a general four-TIP success rate or a claimed confidence interval. [2]
A fast opening does not settle the full match
A separately recorded continuation of seed 31260975 included one AUTO PARK and one late TELEOP PARK. Red finished with 12 total TIPs and 256 points versus Blue's 211. That continuation must not be confused with the separate 13-TIP witnesses in the full-match study. [3]
The useful design hypothesis is to pipeline collection and scoring around the changing HIVE face instead of making both robots wait on one serial workflow. Before calling it a competition strategy, repeat it across broader held-out conditions and physically measure intake reliability, travel, mechanism timing and recovery after a missed first volley. [1][2]
Four completions, one strict cutoff
Red seed 31260975, selected-seed scripted fixture. TIP 4 completes about 0.683 seconds before 30.000. A TIP merely started before 30 would not count.
Exact TIP completion times
| TIP | Completed at | Phase |
|---|---|---|
| 1 | 2.542 s | AUTO |
| 2 | 10.200 s | AUTO |
| 3 | 18.858 s | AUTO |
| 4 | 29.317 s | AUTO |
Discovery and evaluation are different samples
The selected set can establish a modeled possibility. The unselected paired comparison is the more relevant evidence about improvement.
| Sample | Trials | Four-TIP outcomes | What it supports |
|---|---|---|---|
| Previously three-TIP replay seeds | 48 | 2 | A best-case sequence exists in this model |
| Unselected paired Red seeds | 12 per arm | 0 in either arm | Higher mean AUTO count in this small comparison |
Source [2].
What this study does not prove
- Two of 48 is conditional on favorable replay selection, not a general success probability.
- Two active allied robots contribute to this opening; do not compare it as a solo AUTO result.
- The model uses idealized sensors, 2% conditional misses, estimated 0.8-second HIVE movement and uncalibrated pickup behavior.
- The pure AUTO fixture and its full-match continuation are separate runs with different end-of-match behavior.
- Four is the best witnessed count in these experiments, not a proven maximum or hardware demonstration.
Where to play BIOBUZZ
BloxBuzz is where you can play BIOBUZZ; the RLE is where our agents train and are evaluated. The RLE does not run in Roblox, and our goal is for it to use the same game engine, rules and physics as BloxBuzz. These results come from the RLE, not from BloxBuzz gameplay, and this study does not verify that the two match.
Opens Roblox in a new tab. Roblox's age requirements and privacy practices apply. BloxBuzz is unofficial and is not affiliated with or endorsed by FIRST or RTX.
Play BloxBuzz on RobloxSources and evidence
Linked JSON files are sanitized research extracts with sample counts, event witnesses and source hashes. Embedded 2D playback uses exact retained frame coordinates and TIP events; the full original actor logs remain privately retained. These extracts are not a complete reproduction package.
- Selected-seed four-TIP experiment
The 48-trial discovery extract preserves both four-TIP event sequences, selection method and SHA-256 of the original experiment file.
- Unselected paired twelve-seed comparison
Per-seed baseline and staged Red counts, safety indicators, means, ten improvements and two ties. Kept separate from the selected discovery set.
- Full-match continuation of seed 31260975
Separately recorded 150-powered-second run: Red 12 TIPs, scores 256-211. The extract includes its source-file hash.
- Official BIOBUZZ Competition Manual
Consult the current rules before adopting a starting pose, collection plan or parking objective.