Skip to main content
Formal Specification

Blind Cow (Austrian blindfold tag) — Formal Specification

Austria · chase-tag · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 2
# Blinde Kuh — AI-optimized specification ## General idea A multi-agent pursuit-evasion game with partial information. One blindfolded agent navigates a 2D bounded grid with zero visual input, relying purely on localized acoustic sound emission signals produced by moving evaders. Upon capturing an evader, the hunter must solve an identity-classification problem based on tactile feature vectors to transition roles. ## Formal model - **Abstraction taken:** Continuous room movement is discretized into a 2D grid M×M (10×10 cells). Visual observation for the hunter agent is completely masked (blackout). Evaders emit directional acoustic signals S(x,y) whose perceived intensity decays inversely with Euclidean distance. - **Player/agent count:** N agents (4≤N≤15), configured as 1 `Hunter` and N−1 `Evaders`. - **Turn model:** Synchronous discrete movement per tick (T=0.5 seconds per tick). ## Formal rules R1. Setup: N agents are placed in grid M×M. One agent is designated `Hunter` at grid center (5,5) with visibility mask =𝟎. Each agent i has a private tactile feature profile 𝐟i∈ℝ4 (height, texture, clothing, accessory). R2. Movement Phase: - Each tick, all agents choose a move action ∈{North,South,East,West,Stay}. - Hunter moves at speed vH=1 cell/tick. Evaders move at speed vE=1 cell/tick. - An evader moving or generating sound emits an acoustic signal Aj heard by Hunter as noisy bearing θˆ∼𝒩(θ,σsound2). R3. Capture: - If Hunter occupies the same grid cell as an Evader j (PH=Pj), Evader j is captured. - Evader j must remain stationary; Hunter receives noisy observation of 𝐟j. R4. Identification Guess: - Hunter executes `guess_identity(agent_id)`. - If guessed ID =j: Hunter wins the round; captured agent j becomes the new `Hunter` for the subsequent round. - If guessed ID ≠j: [gap-fill] Evader j is released with a 3-tick immunity sprint; Hunter remains `Hunter`. R5. Boundary Enforcement: Evaders cannot move outside the M×M grid; boundary collision forces `Stay`. R6. Match Victory: In an episode of Rmax⁡=10 rounds, the agent who spent the fewest cumulative ticks as `Hunter` wins the match. ## Edge cases - **Tie at the end condition:** [gap-fill] If two agents tie with lowest Hunter time, the agent with the higher identification accuracy on first guesses wins. - **A player with no legal move:** Movement into an adjacent legal grid cell or `Stay` is universally valid. - **Repetition / cycles:** Hunter pursuit bounded by episodic structure and hard round cap Rmax⁡=10. - **Simultaneous claims on the same resource or position:** If Hunter enters a cell containing multiple Evaders, the evader with lowest index is captured first. - **Player elimination mid-round; turn order repair:** No eliminations; role swap occurs seamlessly on successful identification. - **Resource exhaustion:** Finite spatial area and bounded time. - **Termination guarantee:** Guaranteed by finite acoustic tracking time and hard round cap Rmax⁡=10. ## State representation - **Persistent state:** Scoreboard of hunter time per agent, round counter r∈[1,10], active hunter ID. - **Per-round state:** Agent grid coordinates (xi,yi), acoustic signal emission buffer, captured agent ID. - **Observable information:** Highly asymmetric. Evaders observe full 2D grid and hunter position. Hunter observes zero visual map, receiving only noisy bearing vectors θˆ and tactile feature vectors upon contact. ## Action space - `move(direction)`: Enum ∈{N,S,E,W,Stay}. - `emit_sound()`: (Evaders) Deliberate lure signal. - `guess_identity(id)`: (Hunter) Select candidate agent ID. ## Playability notes - **Single-player-vs-AI:** Excellent as an asymmetric audio-spatial puzzle game. - **Group games (n>4):** Naturally operates as a multi-agent hiding and tracking playground. - **Physical-skill games:** Blindfold orientation, auditory triangulation, and tactile identification are modeled via sensor noise and feature matching. - **What is lost in the adaptation:** The physical cloth over eyes, stumbling fear of bumping into furniture, and breathless whispers. ## Implementation hints - Implement grid as a 10×10 2D integer array with distance-decay acoustic broadcast function. - Suggested first milestone: Implement the 2D grid pursuit and capture mechanic with noisy bearing hints before building the tactile identification classifier.

Download as JSON

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

Download JSON
Cite this page

Blind Cow (Austrian blindfold tag) (Austria). (n.d.). Games of Countries. https://gamesofcountries.com/games/at-austria/blinde-kuh

@misc{gamesofcountries-at-austria-blinde-kuh,
  title = {Blind Cow (Austrian blindfold tag) (Austria)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/at-austria/blinde-kuh}
}