Title: RFT Observer Agent Console — update: Predator collision-avoidance tab added (with graphs + logs)
Hi all — quick update on my Space: Agents Console - a Hugging Face Space by RFTSystems
I’ve just updated it with a new Predator (collision-avoidance) agent tab.
What this Space is: a transparent, reproducible test harness for a decision-timing framework I call Rendered Frame Theory (RFT). The core idea is simple: timing matters. I model uncertainty explicitly, adapt an effective timing factor (τ_eff), and use a gate to decide when to commit an action vs wait.
Important clarity:
- I’m calling it an Observer Agent because “observer” here means a decision mechanism (uncertainty → τ_eff → gate → commit/wait).
- I’m not making a machine consciousness claim in this Space.
What’s now inside:
- NEO alerting (noisy tracking + gating)
- Satellite jitter reduction (duty/chatter reduction + residual tracking)
- Starship-style landing harness (simplified timing-control under wind/thrust disturbances)
- Predator collision-avoidance agent (avoid collisions / near-misses under dynamic pursuit)
- Benchmarks (baseline vs RFT, same seed) + plots + CSV logs for every run
What I want from the community (engineering feedback):
- Are my baselines fair and clearly defined?
- Does the Predator tab read as a clean control/decision problem (not a “story”)?
- What metrics should I add for Predator? (collision count, min separation, path efficiency, control effort, latency, etc.)
- If any part reads like over-claiming, tell me and I’ll tighten it.
If you run it and something looks off, please paste your seed + settings and I’ll reproduce it from the logs.