Task 3: add Bio Player (pointwise-activation ESN, 7-slot reflex-only) — 14.40 ± 0.49

[GuanchunLi]

Summary

Adds braincraft/env3_player_bio.py, a minimal handcrafted pointwise-activation Echo State Network controller for Task 3, together with a full model description in braincraft/env3_player_bio.md.

No learning / no training loop — the weight matrices are fixed at build time and the single yield in bio_player() returns the frozen model.

Approach

Same pointwise-ESN skeleton as the env1/env2 bio players — every hidden activation is a scalar function of its own preactivation, all cross-neuron logic lives in the connectivity matrices — but the env3 variant is deliberately stripped down:

• 7 of 1000 hidden slots are active (slots 0–6); the rest are dead.
• Colour and energy inputs are not read. The bot runs the outer corridor as a reflex wall-follower and picks up whatever source it happens to cross, exploiting the known Task 3 outer-ring regularity.
• The circuit is a reflex wall-follower — left/right proximity, left/right safety, hit — plus an unsigned front-block that turns the bot (CCW) when both centre proximity taps exceed 1.4.
• Six non-zero Wout entries; the dtheta slot mirrors the readout row so dtheta(t+1) = clip(O(t), ±5°).

This is the minimal variant of the bio-player family.

Assumed performance

Running python braincraft/env3_player_bio.py (seed 12345, 10 runs)