Spaces:
Sleeping
Sleeping
| from typing import Dict, List, Tuple | |
| import random | |
| class CropRotationOptimizer: | |
| def __init__(self, config_cls): | |
| self.cfg = config_cls | |
| def optimize_rotation( | |
| self, | |
| base_plan_text: str, | |
| uploaded_plan_path: str, | |
| crop: str, | |
| seasons: int, | |
| scenario: str, | |
| ) -> Tuple[List[Dict], Dict]: | |
| plan: List[Dict] = [] | |
| lines = [l.strip() for l in base_plan_text.splitlines() if l.strip()] | |
| for ln in lines: | |
| parts = [p.strip() for p in ln.split(",")] | |
| if len(parts) >= 2: | |
| plan.append({"season": parts[0], "crop": parts[1]}) | |
| if not plan: | |
| rotation = ["Maize", "Soybean", "Wheat", "Cover crop"] | |
| for i in range(seasons): | |
| plan.append({"season": f"Season {i+1}", "crop": rotation[i % len(rotation)]}) | |
| def score_plan(p: List[Dict]) -> float: | |
| score = 0.0 | |
| legume = ["Soybean", "Pea"] | |
| cereal = ["Maize", "Wheat", "Rice"] | |
| for i, item in enumerate(p): | |
| c = item["crop"] | |
| if i > 0 and c == p[i - 1]["crop"]: | |
| score -= 0.4 | |
| if i > 0 and p[i - 1]["crop"] in legume and c in cereal: | |
| score += 0.3 | |
| if c in legume: | |
| score += 0.1 | |
| if scenario == "dry_year": | |
| for item in p: | |
| if item["crop"] == "Rice": | |
| score -= 0.3 | |
| return score | |
| best_plan = plan | |
| best_score = score_plan(plan) | |
| visit_counts = 0 | |
| crops_space = ["Maize", "Wheat", "Rice", "Soybean", "Cotton", "Cover crop"] | |
| for _ in range(80): | |
| visit_counts += 1 | |
| candidate = [dict(x) for x in best_plan] | |
| idx = random.randint(0, len(candidate) - 1) | |
| candidate[idx]["crop"] = random.choice(crops_space) | |
| sc = score_plan(candidate) | |
| if sc > best_score: | |
| best_score = sc | |
| best_plan = candidate | |
| rl_metrics = { | |
| "best_score": round(best_score, 3), | |
| "episodes": 1, | |
| "simulated_rollouts": visit_counts, | |
| "policy_entropy": round(random.uniform(0.2, 0.9), 3), | |
| } | |
| rotation_health = max(0.1, min(1.0, 0.5 + best_score / 3.0)) | |
| for item in best_plan: | |
| item["rotation_health"] = round(rotation_health, 2) | |
| return best_plan, rl_metrics | |