Spaces:
Sleeping
Sleeping
File size: 10,588 Bytes
b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 a099b30 b6f80c5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 | """
Task graders for the Autonomous Traffic Control OpenEnv environment.
Defines three tasks of increasing difficulty:
1. basic_flow – baseline throughput optimisation (Easy)
2. emergency_priority – emergency vehicle management + throughput (Medium)
3. dynamic_scenarios – surge-traffic + emergencies under hard constraints (Hard)
Each grader returns a GradeResult(score, metrics, feedback) with 0–1 score.
Scoring dimensions:
- Throughput : vehicles cleared per step
- Efficiency : low total waiting time
- Emergency rate : emergency vehicles cleared per step
- Emergency delay : average delay per emergency vehicle
- Adaptability : not over-switching phases
- Consistency : steady throughput (low variance) — BONUS
- Queue balance : not letting one direction starve — BONUS
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import Any, Dict
@dataclass
class GradeResult:
"""Standardised grading result."""
score: float # strictly in (0.001, 0.999)
metrics: Dict[str, Any] = field(default_factory=dict)
feedback: str = ""
def _clamp(score: float) -> float:
"""Ensure score is strictly between 0 and 1 (never 0.0 or 1.0 exactly)."""
return round(max(0.001, min(0.999, score)), 4)
# ---------------------------------------------------------------------------
# Public entry point
# ---------------------------------------------------------------------------
def grade(
task_id: str,
*,
total_vehicles_passed: int = 0,
total_emergency_passed: int = 0,
total_waiting_time: float = 0.0,
total_collisions: int = 0,
total_emergency_delay: float = 0.0,
total_phase_changes: int = 0,
step_count: int = 1,
) -> GradeResult:
"""Route to the appropriate task grader."""
graders = {
"basic_flow": _grade_basic_flow,
"emergency_priority": _grade_emergency_priority,
"dynamic_scenarios": _grade_dynamic_scenarios,
}
if task_id not in graders:
return GradeResult(
score=0.001,
feedback=f"Unknown task_id '{task_id}'. Valid: {list(graders.keys())}",
)
return graders[task_id](
total_vehicles_passed=total_vehicles_passed,
total_emergency_passed=total_emergency_passed,
total_waiting_time=total_waiting_time,
total_collisions=total_collisions,
total_emergency_delay=total_emergency_delay,
total_phase_changes=total_phase_changes,
step_count=max(step_count, 1),
)
# ---------------------------------------------------------------------------
# Task 1 – Basic Flow (Easy)
# ---------------------------------------------------------------------------
_BASIC_FLOW_TARGET_THROUGHPUT_PER_STEP = 1.8 # vehicles/step considered "perfect"
def _grade_basic_flow(
*,
total_vehicles_passed: int,
total_waiting_time: float,
total_collisions: int,
total_phase_changes: int,
step_count: int,
**_ignored,
) -> GradeResult:
throughput_per_step = total_vehicles_passed / step_count
throughput_score = min(throughput_per_step / _BASIC_FLOW_TARGET_THROUGHPUT_PER_STEP, 1.0)
efficiency_score = 1.0 / (1.0 + total_waiting_time / max(step_count, 1) * 0.1)
collision_penalty = 0.8 if total_collisions > 0 else 0.0
# BONUS: Queue balance — penalize excessive phase switching (shows instability)
switch_rate = total_phase_changes / max(step_count, 1)
stability_bonus = max(0.0, 0.05 * (1.0 - min(switch_rate * 4, 1.0)))
raw = throughput_score * 0.6 + efficiency_score * 0.4 + stability_bonus
score = max(0.0, raw - collision_penalty)
return GradeResult(
score=_clamp(score),
metrics={
"throughput_per_step": round(throughput_per_step, 3),
"throughput_score": round(throughput_score, 4),
"efficiency_score": round(efficiency_score, 4),
"stability_bonus": round(stability_bonus, 4),
"total_collisions": total_collisions,
"collision_penalty": collision_penalty,
},
feedback=(
f"Throughput {throughput_per_step:.2f} veh/step "
f"(target {_BASIC_FLOW_TARGET_THROUGHPUT_PER_STEP}). "
f"Phase switches: {total_phase_changes} ({switch_rate:.2f}/step). "
+ ("⚠ Collision penalty applied!" if total_collisions else "No collisions ✓.")
),
)
# ---------------------------------------------------------------------------
# Task 2 – Emergency Priority (Medium)
# ---------------------------------------------------------------------------
_EMERG_TARGET_DELAY_PER_VEHICLE = 3.0 # steps/emergency vehicle
def _grade_emergency_priority(
*,
total_vehicles_passed: int,
total_emergency_passed: int,
total_waiting_time: float,
total_collisions: int,
total_emergency_delay: float,
step_count: int,
**_ignored,
) -> GradeResult:
throughput_per_step = total_vehicles_passed / step_count
throughput_score = min(throughput_per_step / 1.5, 1.0)
# Emergency throughput score: 1.0 if ≥ 1 emergency vehicle cleared per 20 steps
em_rate = total_emergency_passed / step_count
em_rate_score = min(em_rate / (1.0 / 20.0), 1.0)
# Emergency delay score
if total_emergency_passed > 0:
avg_delay = total_emergency_delay / total_emergency_passed
delay_score = max(0.0, 1.0 - avg_delay / (_EMERG_TARGET_DELAY_PER_VEHICLE * 4))
else:
delay_score = 0.5
efficiency_score = 1.0 / (1.0 + total_waiting_time / max(step_count, 1) * 0.05)
collision_penalty = 0.85 if total_collisions > 0 else 0.0
# BONUS: emergency response quality
response_bonus = 0.0
if total_emergency_passed > 0:
avg_em_delay = total_emergency_delay / total_emergency_passed
if avg_em_delay < 2.0:
response_bonus = 0.05 # exceptional response time
elif avg_em_delay < 4.0:
response_bonus = 0.02
raw = (throughput_score * 0.30 + em_rate_score * 0.35 +
delay_score * 0.20 + efficiency_score * 0.15 + response_bonus)
score = max(0.0, raw - collision_penalty)
avg_delay_str = (
f"{total_emergency_delay / total_emergency_passed:.1f} steps"
if total_emergency_passed else "N/A"
)
return GradeResult(
score=_clamp(score),
metrics={
"throughput_per_step": round(throughput_per_step, 3),
"throughput_score": round(throughput_score, 4),
"emergency_rate_score": round(em_rate_score, 4),
"emergency_delay_score": round(delay_score, 4),
"efficiency_score": round(efficiency_score, 4),
"response_bonus": round(response_bonus, 4),
"total_emergency_passed": total_emergency_passed,
"avg_emergency_delay_steps": avg_delay_str,
"total_collisions": total_collisions,
},
feedback=(
f"Cleared {total_emergency_passed} emergency vehicles "
f"(avg delay {avg_delay_str}). "
f"Throughput {throughput_per_step:.2f} veh/step. "
+ (f"🏆 Fast response bonus +{response_bonus:.0%}! " if response_bonus > 0 else "")
+ ("⚠ Collision!" if total_collisions else "No collisions ✓.")
),
)
# ---------------------------------------------------------------------------
# Task 3 – Dynamic Scenarios (Hard)
# ---------------------------------------------------------------------------
def _grade_dynamic_scenarios(
*,
total_vehicles_passed: int,
total_emergency_passed: int,
total_waiting_time: float,
total_collisions: int,
total_emergency_delay: float,
total_phase_changes: int,
step_count: int,
**_ignored,
) -> GradeResult:
throughput_per_step = total_vehicles_passed / step_count
throughput_score = min(throughput_per_step / 2.0, 1.0)
em_rate = total_emergency_passed / step_count
em_rate_score = min(em_rate / (1.0 / 15.0), 1.0)
if total_emergency_passed > 0:
avg_delay = total_emergency_delay / total_emergency_passed
delay_score = max(0.0, 1.0 - avg_delay / 5.0)
else:
delay_score = 0.0
efficiency_score = 1.0 / (1.0 + total_waiting_time / max(step_count, 1) * 0.08)
adaptability_score = 1.0 / (1.0 + total_phase_changes / max(step_count, 1) * 0.5)
collision_penalty = 0.9 if total_collisions > 0 else 0.0
# BONUS: queue balance + surge resilience
surge_bonus = 0.0
if total_vehicles_passed > step_count * 1.5:
surge_bonus = 0.03 # handled high traffic well
if total_emergency_passed > 0 and total_collisions == 0:
surge_bonus += 0.02 # survived with zero collisions
raw = (throughput_score * 0.25 + em_rate_score * 0.30 +
delay_score * 0.20 + efficiency_score * 0.15 +
adaptability_score * 0.10 + surge_bonus)
score = max(0.0, raw - collision_penalty)
return GradeResult(
score=_clamp(score),
metrics={
"throughput_per_step": round(throughput_per_step, 3),
"throughput_score": round(throughput_score, 4),
"emergency_rate_score": round(em_rate_score, 4),
"emergency_delay_score": round(delay_score, 4),
"efficiency_score": round(efficiency_score, 4),
"adaptability_score": round(adaptability_score, 4),
"surge_bonus": round(surge_bonus, 4),
"total_collisions": total_collisions,
"total_phase_changes": total_phase_changes,
},
feedback=(
f"Dynamic task: throughput {throughput_per_step:.2f} veh/step, "
f"{total_emergency_passed} emergencies cleared, "
f"{total_phase_changes} phase changes over {step_count} steps. "
+ (f"🏆 Surge resilience bonus +{surge_bonus:.0%}! " if surge_bonus > 0 else "")
+ ("⚠ Collision!" if total_collisions else "No collisions ✓.")
),
)
|