Skip to main content
Formal Specification

Kadazan-Dusun Bamboo Stilt Racing — Formal Specification

Malaysia · strength-endurance · Confidence: high

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: abstractedGap-fills: 3
# Rampanau — AI-optimized specification ## General idea A simultaneous-turn locomotion and dynamic balance racing simulation based on traditional Kadazan-Dusun bamboo stilt racing. Agents coordinate alternating left-right stilt strides along a 40-meter track, managing stride cadence, forward propulsion force, and pitch-roll stability against escalating biomechanical fatigue to achieve minimum race time without exceeding fall thresholds. ## Formal model - **Abstraction taken:** Stilt locomotion is modeled as a coupled inverted-pendulum dynamical system on a 1D track. The continuous state per agent i is vector Si=(xi,vi,θpitch,θroll,fall_count,Ffatigue), where xi∈[0,Dfinish] (Dfinish=40.0 m), vi is sprint speed, and θ angles track stilt tilt relative to vertical. - **Player/agent count:** 2 to 8 agents racing simultaneously in dedicated lanes. - **Turn model:** Simultaneous-turn discrete-time simulation (Δt=0.1 s). Each tick, all agents choose stride cadence and balance adjustment impulses. ## Formal rules R1. State tracking & track layout: - Track length: Dfinish=40.0 m. - Each agent i occupies a dedicated lane of width W=2.0 m. - Stilt elevation: Footrest height hfoot=0.5 m. R2. Locomotion & balance update per tick: - Agent commits stride cadence ucadence∈[0,umax] and balance corrective force ubal∈[−Bmax,Bmax]. - Velocity update: vi(t+1)=vi(t)+(α⋅ucadence−β⋅vi(t)−γ⋅Ffatigue)Δt - Balance drift: θpitch(t+1)=θpitch(t)+(kp⋅vi(t)+ubal)Δt+ϵterrain where ϵterrain∼𝒩(0,σground2). - Position update: xi(t+1)=xi(t)+vi(t)Δt. - Fatigue accumulation: Ffatigue(t+1)=min⁡(1.0,Ffatigue(t)+δf⋅ucadenceΔt). R3. Fall and dismount penalty: - If |θpitch|>θcrit (25∘), agent loses balance and dismounts (*jatuh*). - Status transitions to `DISMOUNTED`. - Counter updates: fall_counti←fall_counti+1. - Mandatory remount penalty: Velocity drops to vi←0, and agent must remain stationary for Tremount=3.0 s (30 ticks) to reset balance angles before resuming forward strides ( [gap-fill]). R4. Disqualification condition: - If fall_counti≥3, agent i is immediately `DISQUALIFIED` and removed from the active race. R5. Victory condition: - The first agent whose position reaches xi≥Dfinish while upright and mounted wins the heat. R6. [gap-fill] Hard heat time ceiling: If no agent completes the course within Tmax=60.0 s (600 ticks), the non-disqualified agent with the greatest distance xi is declared the winner. R7. [gap-fill] Multi-heat tournament scoring: In multi-heat tournaments, agents accumulate championship points based on finish order per heat (1st: 10 pts, 2nd: 6 pts, 3rd: 4 pts, 4th: 2 pts). ## Edge cases - Tie at the end condition: Sub-tick crossing time interpolation; if identical, broken by lower cumulative fall_count. - A player with no legal move: Not applicable; agents can always select minimal cadence (u=0) for balance stabilization. - Repetition / cycles: Impossible due to monotonic forward progress or hard 600-tick cutoff. - Simultaneous claims on the same resource or position: Agents run in distinct, non-intersecting parallel lanes. - Player elimination mid-round; turn order repair: Disqualified agents are flagged as `DISQUALIFIED` and halt; remaining racers continue to the finish line. - Resource exhaustion (pieces, tokens, space, time): Track length is fixed at 40 meters; simulation terminates within 600 ticks. - Termination guarantee: Monotonic progress mechanics and the hard 600-tick timeout provably ensure termination. ## State representation - Persistent state: `track_length`, `lane_count`, `race_time`, `rankings`. - Per-agent state: `(x, v, pitch, roll, fall_count, fatigue, status, remount_ticks_left)`. - Observable information: Fully observable by all agents. ## Action space - `step_stilts(cadence, balance_impulse)`: Continuous 2D control action where cadence∈[0,umax] and balance_impulse∈[−Bmax,Bmax]. ## Playability notes - Single-player-vs-AI: Excellent arcade racing game against AI pace controllers with varying risk profiles. - Group games (n>4): Natively supports 2 to 8 parallel racing agents. - Physical-skill games: Dynamic balance and leg stamina are cleanly modeled via inverted pendulum kinematics and non-linear fatigue dampening. - What is lost in the adaptation: The visceral tactile sensation of standing on 0.5-meter elevated bamboo footholds and balancing via upper-body arm grips. ## Implementation hints Model as a continuous control benchmark (Gymnasium environment). Action space: Box(2). Observation space: Box(6). First milestone: implement a baseline PID controller that successfully completes the 40-meter course without falling before introducing terrain noise.

Download as JSON

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

Download JSON
Cite this page

Kadazan-Dusun Bamboo Stilt Racing (Malaysia). (n.d.). Games of Countries. https://gamesofcountries.com/games/my-malaysia/rampanau

@misc{gamesofcountries-my-malaysia-rampanau,
  title = {Kadazan-Dusun Bamboo Stilt Racing (Malaysia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/my-malaysia/rampanau}
}