Skip to content

TEST_CASES.md

Latest commit

 

History

History
322 lines (259 loc) · 6.38 KB

TEST_CASES.md

File metadata and controls

322 lines (259 loc) · 6.38 KB

Enhanced Greedy CPU AI - Test Cases

Scoring Formula

score = (data_collected × 100) - cluster_distance
cluster_distance = avg Euclidean distance to 3 closest remaining data points

Category 1: Cluster vs Isolated Data Advantage

TC-CA-001: Dense Cluster vs Single Distant Data

Grid (6×6):

@ . . . . D
. . . . . .
. D D . . .
. D . . . .
. . . . . .
. . . . . H

Setup:

  • Packet @ at (0,0)
  • Data at: D1(5,0), D2(1,2), D3(2,2), D4(1,3)
  • Cluster: D2,D3,D4 form tight group at ~(1.3, 2.3)

Direction Analysis:

  • East (→): Slides to (5,0), collects D1

    • Data collected: 1
    • Remaining: D2,D3,D4
    • Cluster dist: avg((2.24, 3.16, 4.24)) = 3.21
    • Score: 100 - 3.21 = 96.79

  • Southeast (↘): Slides to (1,2), collects D2

    • Data collected: 1
    • Remaining: D1,D3,D4
    • Cluster dist: avg((1.0, 1.41, 4.58)) = 2.33
    • Score: 100 - 2.33 = 97.67 ✓

Expected:

  • New AI: Chooses Southeast (97.67) - positions near remaining cluster
  • Old AI: Would choose East (nearest single data)
  • Pass Criteria: CPU moves Southeast, demonstrating cluster awareness

TC-CA-002: Two Clusters - Near vs Far

Grid (8×8):

@ . D D . . . .
. . D . . . . .
. . . . . . . .
. . . . . . D D
. . . . . . D .
. . . . . . . .
. . . . . . . .
. . . . . . . H

Setup:

  • Packet @ at (0,0)
  • Cluster A: D1(2,0), D2(3,0), D3(2,1) at ~(2.3, 0.3)
  • Cluster B: D4(6,3), D5(7,3), D6(6,4) at ~(6.3, 3.3)

Direction Analysis:

  • East: Slides to (2,0), collects D1
    • Remaining: D2,D3 (near), D4,D5,D6 (far)
    • 3 closest: D2(1.0), D3(1.41), D4(6.08)
    • Cluster dist: 2.83
    • Score: 100 - 2.83 = 97.17 ✓

Expected:

  • New AI: Navigates strategically between clusters
  • Pass Criteria: CPU evaluates both clusters and picks optimal positioning

TC-CA-003: Multiple Data in Path

Grid (6×6):

@ D D D . .
. . . . . .
. . . . . .
. . . D D D
. . . . . .
. . . . . H

Setup:

  • Packet @ at (0,0)
  • Top row: D1(1,0), D2(2,0), D3(3,0)
  • Bottom row: D4(3,3), D5(4,3), D6(5,3)

Direction Analysis:

  • East: Slides to (3,0), collects D1,D2,D3 (3 data!)
    • Data collected: 3
    • Remaining: D4,D5,D6
    • Cluster dist to D4,D5,D6: avg((3.0, 4.0, 5.0)) = 4.0
    • Score: 300 - 4.0 = 296.0 ✓

Expected:

  • New AI: Chooses East - immediate collection outweighs positioning
  • Pass Criteria: CPU correctly prioritizes multiple immediate data

Category 2: Memoization Hits

TC-MEM-001: Same End Position from Two Directions

Grid (6×6):

. . @ . . .
. . # # # .
. . # D # .
. . # # # .
. . . . . .
. . . . . H

Setup:

  • Packet @ at (2,0)
  • Data D at (3,2) (surrounded by firewalls)

Memoization:

  • Multiple directions might simulate ending at same (x,y) with same remaining data
  • Pass Criteria:
    • Cluster distance calculated only once for duplicate positions
    • Cache hit logs in debug mode

TC-MEM-002: Symmetric Board Layout

Grid (6×6):

. D . . D .
. . . . . .
@ . . . . .
. . . . . .
. . . . . .
. D . . D H

Setup:

  • Packet @ at (0,2)
  • Symmetric data: (1,0), (4,0), (1,5), (4,5)

Memoization:

  • Different positions but potentially same data state
  • Pass Criteria: Memo cache differentiates by position despite data symmetry

Category 3: Edge Cases (0-2 Data Points Left)

TC-EDGE-001: Single Data Point Remaining

Grid (6×6):

@ . . . . .
. . . . . .
. . . . D .
. . . . . .
. . . . . .
. . . . . H

Setup:

  • Packet @ at (0,0)
  • Only 1 data D at (4,2)

Expected Behavior:

  • Cluster distance: With only 1 data point, use distance to that point
  • Southeast: Ends at (4,2), collects D, Score = 100 ✓

Pass Criteria:

  • CPU handles single-data case without array index errors
  • Correctly navigates to collect last data point

TC-EDGE-002: Two Data Points Remaining

Grid (6×6):

@ . . D . .
. . . . . .
. . . . . .
. D . . . .
. . . . . .
. . . . . H

Setup:

  • Packet @ at (0,0)
  • Two data: D1(3,0), D2(1,3)

Expected Behavior:

  • Cluster distance: Avg of 2 closest (both used)
  • Pass Criteria: Handles 2-data case, correct decision

TC-EDGE-003: Zero Data Remaining

Grid (6×6):

@ . . . . .
. # . . . .
. # . . . .
. # . . . .
. # . . . .
. . . . . H

Setup:

  • All data already collected
  • Hub H at (5,5)

Expected Behavior:

  • Cluster distance: No data left → cluster_dist = 0 (neutral)
  • Pass Criteria: No divide-by-zero errors, valid move toward hub

Category 4: Virus/Firewall Interference

TC-SAFE-001: Cluster Beyond Virus

Grid (6×6):

@ . . V D D
. . . . D .
. . . . . .
. . . . . .
. . . . . .
. . . . . H

Setup:

  • Packet @ at (0,0)
  • Virus V at (3,0)
  • Data cluster: D1(4,0), D2(5,0), D3(4,1)

Expected:

  • East: Slides to (3,0) - HITS VIRUS ❌ (filtered out)
  • Pass Criteria: CPU never chooses direction leading to virus

TC-SAFE-002: Firewall Blocks Optimal Path

Grid (8×8):

@ # D D D D
. # . . . .
. # . . . .
. # . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . H

Setup:

  • Firewall column # at x=1
  • Dense data cluster at (2,0), (3,0), (4,0), (5,0)

Expected:

  • Pass Criteria: Firewall collision detection prevents invalid moves

Category 5: Blocking / No Valid Moves

TC-BLOCK-001: Completely Surrounded

Grid (6×6):

. . . . . .
. # # # . .
. # @ # . .
. # # # . D
. . . . . .
. . . . . H

Setup:

  • Packet @ at (2,2)
  • Surrounded by firewalls

Expected Behavior:

  • All 8 directions: Blocked by firewall
  • CPU Response: Returns null (no valid move)
  • Pass Criteria: No crash, graceful "no valid move" handling

TC-BLOCK-002: All Directions Lead to Virus

Grid (6×6):

. V V V . .
V V @ V V D
. V V V . .
. . . . . .
. . . . . .
. . . . . H

Setup:

  • Surrounded by viruses

Expected Behavior:

  • Pass Criteria: CPU doesn't commit suicide, returns null gracefully

Expected Improvements

Scenario Old Greedy New Cluster-Aware Improvement
Dense cluster Picks nearest Positions near cluster +15% win rate
Multiple isolated Random-ish Strategic order +20% efficiency
End-game (2-3 left) Greedy only Optimal positioning +10% win rate
Complex boards O(n²) calc O(n log n) + memo 2x faster

Total Test Cases: 12 across 5 categories