Registrations & Environment Online

Beyond
the dunes.

The flagship autonomous robotics challenge. Match noisy terrain clues, track visual beacons, and reverse-engineer telemetry inside ROS 2 & Gazebo.

Clearpath Jackal Autonomous Rover in Desert Dunes
Clearpath Jackal J100ROS 2 Jazzy
import rclpy
from rclpy.node import Node
from geometry_msgs.msg import Twist
from beyond_the_dunes.msg import ElevationPatch, VisualJolt

class AutonomousRover(Node):
    def __init__(self):
        super().__init__('autonomous_rover')
        self.cmd_pub = self.create_publisher(Twist, '/cmd_vel', 10)
        self.dem_sub = self.create_subscription(
            ElevationPatch, '/arena/clue', self.on_clue_received, 10
        )
        self.get_logger().info('Rover online. Listening for 30×30m DEM clues...')

    def on_clue_received(self, msg: ElevationPatch):
        # Match local heightmap fragment against 400×400m global terrain
        coords = self.match_terrain_fragment(msg.height_matrix)
        self.navigate_to_target(coords)
        self.trigger_beacon(msg.target_aruco_id)
Gazebo Live Telemetry Feed
Physics: ODE 1000 Hz · Sensor Rate: 30 FPS
[00:04.21] [DEM MATCH]: Fragment P̃_1 correlated at (x: 184.2m, y: -42.8m) — 99.1% match
[00:09.65] [NAV]: Waypoint trajectory engaged. Linear velocity: 1.2 m/s
[00:15.82] [BEACON]: ArUco ID #42 detected at 3.4m. Rover heading aligned.
[00:17.04] [REFEREE]: Checkpoint 1 verified. Visual Jolt HUD signature [+, 0] recorded.
01
₹1,00,000+

Total Prize Pool

Awarded to top 3 autonomous teams

02
400×400m

Global Elevation Grid

OpenTopography realistic terrain

03
05 Stages

Mission Checkpoints

Gated sequentially by referee

04
10⁶ Ticks

PRNG State Space

Pruned via multi-bar visual HUD

05
120 FPS

Simulation Runtime

Gazebo Harmonic on dedicated GPUs

06
2 to 4

Team Size

Verified college participants

Story-Driven Navigation Map

400×400m Terrain Expedition Route

Trace the progressive path of the Clearpath Jackal rover across all five sequential checkpoints in Gazebo Harmonic.

GRID NORTH (400×400m)
SPAWN (0,0)CP1CP2CP3CP4CP5
Active Target: CP1X: 114.2m, Y: 88.5m
STAGE 1 OF 5+42.8m above datum

Checkpoint 1: Sand Crest Ridge

Noisy DEM Correlation

Receive noisy 30×30m patch via /btd/checkpoint_clue and correlate against 400×400m global DEM matrix to lock target coordinates.

ROS 2 Channel:/btd/checkpoint_clue

Normalized 2D Cross-Correlation (NCC) threshold r ≥ 0.90 required to filter noise artifacts and lock waypoint.

Technical Architecture Interactive Demo

Simulate the Four Core Algorithms

Test and observe the exact robotics principles participants encounter during an official Beyond the Dunes mission run.

Matrix Canvas: 400×400m Global GridNCC Pearson: r = 0.632
30×30m PatchNoise: 15% Gaussian
SEARCHING...
Ground Truth Waypoint: (X: 18, Y: 18)Tolerance: r ≥ 0.880
DEM PATCH CORRELATOR

2D Normalized Cross-Correlation

Referees provide a noisy 30×30m elevation fragment. Slide the patch along the matrix axis to maximize the Pearson correlation coefficient.

Patch Coordinate Sweep:12m
Simulated Sensor Noise (σ):15%
Visual Jolt Protocol • Dart Spark HUD

10-Slot Pixel-Art Telemetry Bar

At each checkpoint, the rover observes the 10-slot visual HUD bar. Slots expose dynamic UP (+), STILL (0), or DOWN (-) states used to crack the master PRNG seed.

B1
• STILL
B2
+ UP
B3
- DOWN
B4
• STILL
B5
- DOWN
B6
• STILL
B7
+ UP
B8
- DOWN
B9
• STILL
B10
+ UP
B₂ & B₅: Primary Exposed SignalsB₁..B₁₀: Click any slot to toggle UP/STILL/DOWN
Click to interact & test PRNG reduction
Dart Spark Formula
S₀ = 48 × F + C

Food parameter F ∈ [0..19]. Candidate seed S₀ branches across 10⁶ ticks.

Selected Slot Telemetry
Bar B2: UPcode: -1

Level: 9/10 blocks illuminated.

Target PRNG Seed Candidate
0x0002BB

Filtered across 5 sequential checkpoints to 1 unique master tick.

Architected for Autonomous Systems

Engineered for teams who think in algorithms.

Zero legacy bloat. Fast turnarounds. Built on modern robotics benchmarks.

Elevation Analysis

400×400m Terrain Correlation

Every world features authentic OpenTopography elevation models. Correlate noisy 30×30m patch fragments against the global elevation matrix to recover precise expedition coordinates.

Sub-meter accuracy required for waypoint lock
Pure Logic

Zero SLAM Overhead

Ground-truth robot 6-DOF pose is streamed directly from the referee. Focus 100% of your compute on navigation strategy, computer vision, and state recovery.

geometry_msgs/PoseStamped
Simulation Stack

Gazebo Harmonic

Native ROS 2 Jazzy execution with ODE physics, high-fidelity GPU lidar, camera sensors, and standardized vehicle twist commands.

120 FPS Dedicated GPU Runner
Evaluation Engine

Automated GPU Runner & Live Standings

Submit controller ZIP packages via CLI or Web. Code runs inside an isolated container with private mission seeds, yielding verified score sheets and public leaderboard rankings.

Ranking Order:Progression (1st) → Score (2nd) → Runtime (Tiebreaker)
Live Standings Preview

Official Qualifier Leaderboard

Ranked strictly by highest progression, then task score, with sim time as tiebreaker.

View Full Leaderboard
RankTeam & AffiliationProgressionTask ScoreSim Time
1
Dust Circuit
IIIT Allahabad
5 / 5 CP1,485 pts8m 42s
2
Sand Nomads
IIT Roorkee
5 / 5 CP1,420 pts9m 15s
3
Dune Drifters
BITS Pilani
5 / 5 CP1,390 pts9m 58s
4
Vector Ridge
IIT Kanpur
4 / 5 CP1,180 pts11m 02s

How to Participate

From zero to qualifier standing.

01

Register with College Email

Sign in with your institutional .edu or .ac.in email. Set up a secure password to maintain persistent browser access.

02

Assemble Your Rover Team

Form a team of 2 to 4 verified college students. Generate your unique 6-character invite code and invite your teammates.

03

Test & Submit Controllers

Download the baseline ROS 2 Jazzy workspace. Test against open practice worlds, then submit your bundle to the GPU queue.

The Starting Line Is Open

Ready to push the boundary?

Join universities from across the nation in the ultimate test of autonomous navigation, computer vision, and reverse engineering.