Flappy-Bird-like Autopilot With Coins as Jump Budget

Problem: Flappy-Bird-like Autopilot With Coins as Jump Budget You are given a simplified side-scrolling jumping game. You must implement a policy that decides whether to jump at each time step (tick). Game Rules The agent moves vertically: Each tick it drops due to gravity (exact update handled by the evaluator). If you choose jump on a tick, the agent moves upward (exact impulse handled by the evaluator). Valid vertical range is [0, H]: If height < 0 (hits the ground), you lose. If height > H (hits the ceiling), you lose. Coins exist on the map: Touching a coin grants score (e.g., +1) (as defined by the evaluator). Each jump consumes 1 coin (coins are your jump budget). When you run out of coins, you can still choose not to jump; if you later crash because you cannot jump, you still lose. Task Implement a policy that, at every tick, decides whether to jump based on the current state, aiming to: Avoid crashing into the ground or ceiling. Collect as many coins / maximize score when possible. Avoid wasting coins so you don’t lose later due to lack of jump budget. Interface (conceptual) The evaluator calls your policy every tick with a state (e.g., current height, vertical velocity, remaining coins, nearby coin positions, etc.; exact fields depend on the OA template). Your policy returns a boolean: True => jump this tick (consumes 1 coin) False => do not jump Notes You do not implement physics or collision; the evaluator does. You only implement decision logic to pass hidden unit tests. The OA UI provides an Auto-pilot visualization to replay the agent’s behavior. Example (format illustration only) Input: evaluator repeatedly calls policy(state) Output: each call returns jump or no jump This is typically an online control/policy problem rather than a classic offline LeetCode-style algorithm question.