Skip to main content
Formal Specification

Blind Hen — Formal Specification

Lithuania · hide-seek · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: discretizedGap-fills: 9
# Akla višta — AI-optimized specification ## General idea Akla višta is an asymmetric blindfold tag and sensory inference game played on a bounded 2D grid. One designated agent (the "blind hen") operates under complete visual deprivation (FOV=∅) and must localize, pursue, and capture evasive dodgers using only noisy directional sound cues emitted during movement, teasing, and clucking. Upon making physical contact, the hen must successfully deduce the target agent's identity via tactile feature probing to transfer the blindfold role. ## Formal model - **Abstraction taken:** Discretized. Continuous 2D physical space is mapped to an M×M grid graph with orthogonal connectivity. Continuous acoustics are modeled as discrete auditory emission packets characterized by loudness, frequency, and positional decay with Gaussian direction noise [gap-fill]. Tactile recognition is abstracted as a deterministic or probabilistic comparison of discrete agent physical attribute vectors [gap-fill]. - **Player/agent count:** N agents (4≤N≤15): 1 blind hen (H) and N−1 dodgers (D1,…,DN−1). - **Turn model:** Synchronous discrete time ticks (t=1,2,…). All agents submit actions simultaneously; movements resolve concurrently, followed by acoustic propagation and contact checks. ## Formal rules R1. Setup: Agents are placed on an M×M grid (M=10 default) [gap-fill]. H is placed at the center (⌊M/2⌋,⌊M/2⌋) with orientation θ∈{0∘,90∘,180∘,270∘}. Dodgers are spawned randomly on non-center cells. R2. Attributes: Each agent i possesses a static physical feature vector F→i=⟨height,hair_type,cloth_texture⟩∈ℝ3 [gap-fill]. R3. Movement and Actions: Each tick, an agent selects one action: - Dodgers: `move(direction)`, `tease(sound_level)`, `freeze()`. - Blind hen: `step(direction)`, `reach_tag()`, `listen()`. R4. Acoustic Propagation: When a dodger moves or teases, an acoustic packet with power P∈[1,5] is emitted. H receives a noisy observation zˆ=(r+𝒩(0,σr2),ϕ+𝒩(0,σϕ2)) where r is Euclidean distance and ϕ is relative bearing [gap-fill]. If a dodger selects `freeze()`, P=0. R5. Boundary Violation: A dodger stepping outside the M×M perimeter incurs a penalty state: immobilized (`frozen = true`) for 5 consecutive ticks. R6. Capture and Tactile Inspection: - If H executes `reach_tag()` and occupies the same cell or an orthogonally adjacent cell to a dodger Dj, Dj is tagged and enters `inspected` state. - H receives a noisy observation of the captured agent's feature vector Fˆ∼𝒩(F→j,Σtactile). - H must submit an identity hypothesis k∈{1,…,N−1}. R7. Role Transfer: - If k=j (correct guess), role transfer occurs: Dj becomes the new H, the round ends, and a new round begins with Dj centered. - If k≠j (incorrect guess), Dj is released, H is stunned for 2 ticks [gap-fill], and the pursuit continues. R8. Game Limit: A single game session consists of R rounds or a maximum global tick budget Tmax⁡=500 ticks per round [gap-fill]. ## Edge cases - Tie at the end condition: Not applicable — role transfer occurs individually upon the first correct tactile identification; round ends immediately. - A player with no legal move: Not applicable — `freeze()` or `listen()` are always valid null moves if movement is blocked. - Repetition / cycles (and the loop-breaking rule): Dodgers cycling indefinitely in evasion are forced into acoustic exposure by a mandatory tease requirement: any dodger stationary for >10 ticks emits an automatic sound packet [gap-fill]. - Simultaneous claims on the same resource or position: Multiple dodgers may occupy the same grid cell; if H executes `reach_tag()` on a multi-occupancy cell, one dodger is selected uniformly at random for tactile inspection. - Player elimination mid-round; turn order repair: Not applicable — no player is eliminated; active roles rotate dynamically upon successful identification. - Resource exhaustion (pieces, tokens, space, time): Round ends in a draw/timeout if t exceeds Tmax⁡=500 ticks [gap-fill], forcing a role swap via lot draw. - Termination guarantee: Guaranteed by the finite tick budget Tmax⁡ and the non-zero capture probability under bounded grid dimensions and mandatory sound emissions. ## State representation - Persistent state: Grid dimensions M×M, agent role mapping (rolei∈{Hen,Dodger}), round count, cumulative hen-time scores. - Per-turn ephemeral state: Agent positions (xi,yi), velocities, acoustic emission vectors, tag contact status, guess payload k. - Observable information: - Dodgers observe the complete global state (positions of all agents and boundaries). - Blind hen observes only its own coordinate/orientation and incoming auditory observation vectors zˆ; visual grid is completely occluded. ## Action space - `move(dir)`: Step to adjacent cell ∈{North,South,East,West} (all agents). - `tease(level)`: Emit intentional auditory ping with loudness level∈{1,2,3} (dodgers only). - `reach_tag()`: Attempt to grasp adjacent targets in facing direction (hen only). - `identify(agent_id)`: Submit identity guess for captured agent (hen only, in inspection phase). - `freeze()`: Remain stationary, suppressing acoustic emissions (dodgers only). ## Playability notes - Single-player-vs-AI: Highly natural — player can control the blind hen (filtering audio cues through a radar/audio interface) against AI dodgers, or play a dodger evading an algorithmic particle-filter tracker. - Group games (n>4): Reducible to 2 agents (1 hen vs 1 dodger) cleanly, though group dynamics (acoustic clutter, multi-target ambiguity, sacrificial decoying) are reduced. - Physical-skill games: Tactile touch and hearing are formalized as Bayesian classification and Gaussian bearing estimates. - What is lost: The physical sensation of disorientation after spinning, tactile fabric textures, and spontaneous vocal laughter are replaced by mathematical signals. ## Implementation hints State tracking requires a 2D spatial grid and a particle filter or Bayesian belief grid for the blind hen's hidden state estimation. First milestone: Build the 10×10 grid environment with 1 hen and 3 dodgers, implementing sound emission propagation and tag detection without tactile feature noise.

Download as JSON

Download the full game corpus entry for this game as structured JSON.

Download JSON
Cite this page

Blind Hen (Lithuania). (n.d.). Games of Countries. https://gamesofcountries.com/games/lt-lithuania/akla-vista

@misc{gamesofcountries-lt-lithuania-akla-vista,
  title = {Blind Hen (Lithuania)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/lt-lithuania/akla-vista}
}