Skip to content

PROBLEM_STATEMENT.md

Latest commit

 

History

History
110 lines (76 loc) · 3.34 KB

PROBLEM_STATEMENT.md

File metadata and controls

110 lines (76 loc) · 3.34 KB
Raw

Level 2 Demo

Level 2 — Operation Touchdown: Precision Landing

Team Aerial Robotics IITK | Y25 Recruitment Hackathon

1. Context

A drone is descending. You cannot stop it. The landing platform is oscillating sinusoidally below you — it never stops moving. Your only sensor is a downward-facing camera. You must steer the drone laterally so it touches down on the platform centre. No GPS. No position sensor. Everything inferred from pixels.

2. Environment

Simulation

Parameter Value
Sim Duration 35 seconds — fixed descent, cannot be paused
Frame Rate 30 FPS (dt = 0.033 s)
Success Radius Final error ≤ 0.05 m from platform centre
Drone Start (80 px, 180 px) — top-left, far from platform
Altitude 10.0 m → 0.0 m, auto-descending (uncontrollable)
Max Speed 5.0 m/s lateral (simulator enforced)

Landing Platform

Parameter Value
Size 1 m × 1 m
X Motion centre_x + 2.0 × sin(0.9 t) m — SHM, ±2 m amplitude
Y Motion centre_y + 0.6 × sin(0.45 t) m — slow sinusoidal drift

3. Camera

Each frame step_env() returns a flat list of 10 000 grayscale integers (100 × 100 px, row-major).

Field of view: fov_m = 0.30 × altitude metres. Use this to convert pixel offsets to real-world metres.

Region Gray Value Notes
Grass / background ~45–90 Dark green
Platform surface ~200–230 Bright rectangle — the landing pad
Inner square ~0–20 Near-black square at platform centre

4. Your Task

Edit solver.py only. Implement the six TODOs:

TODO Where What
1 SEARCH_SPEED Drone speed (m/s) during search.
2 KP/KI/KD constants PID gains for X and Y axes.
3 detect_platform() Find platform in pixel array. Return (found, cx_norm, cy_norm).
4 PID.update() Implement P + I + D with anti-windup. Return clamped velocity.
5 search_velocity() Design a search pattern to sweep the arena.
6 main() — PID block Convert cx_norm / cy_norm → metres via sim.fov_m. Feed into PID.

Phase 1 — Search

The drone starts top-left; the platform is not visible. Implement search_velocity() to sweep the arena until the platform enters the camera FOV.

Phase 2 — Track & Land

Once detected, use PID to keep the drone centred over the moving platform throughout descent.

err_x_m = cx_norm * (sim.fov_m / 2)
err_y_m = cy_norm * (sim.fov_m / 2)
vx = pid_x.update(err_x_m, dt)
vy = pid_y.update(err_y_m, dt)

5. Scoring

Outcome Points
Final error ≤ 0.20 m 10
Final error ≤ 0.10 m 20
Final error ≤ 0.05 m (SUCCESS) 50
Bonus: error ≤ 0.02 m +20

6. Rules

Allowed

  • Edit solver.py freely — add functions, tune constants, import libraries.
  • Use any pip-installable package (NumPy, OpenCV, SciPy, etc.).

Not Allowed

  • Modify simulator_level2.py.
  • Access sim.plat_x, sim.plat_y, or any internal simulator variable.
  • Hardcode the platform position or trajectory.

Deliverables

Submit solver.py only. Must run with the original unmodified simulator_level2.py in the same directory:

python solver.py