Autonomous UUV Swarm for Undersea Threat Detection & Response
Develop an advanced autonomous undersea vehicle (UUV) swarm coordination system designed to detect, classify, and track unknown underwater objects in challenging, dynamic environments. This challenge emphasizes multi-sensor fusion, real-time decision-making, and intelligent mission planning to effectively address maritime security threats. The system should utilize heterogeneous sensor data (e.g., acoustic, sonar) to form a coherent understanding of the underwater battlespace. A core component involves integrating a large language model, specifically Llama 3.1, to interpret complex scenarios, assess potential threats, and dynamically re-plan UUV trajectories or search patterns based on observed anomalies and overarching mission objectives. Participants will build a simulated environment to test UUV capabilities under various conditions, including communication constraints and environmental noise. The solution should demonstrate robustness, efficient swarm coordination, and effective threat neutralization strategies, deploying edge ML models using BentoML for on-board processing.
What you are building
The core problem, expected build, and operating context for this challenge.
Implement an autonomous undersea vehicle (UUV) swarm coordination system using Llama 3.1 for intelligent mission planning and BentoML for model deployment.
How work is evaluated
The evaluation will focus on the system's ability to autonomously detect, classify, and track simulated undersea threats within a specified time frame, minimizing false positives and maximizing coverage, while leveraging the AI for dynamic decision-making and re-planning.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Implement real-time sensor fusion algorithms for acoustic, sonar, and visual data (simulated).
Design a multi-UUV coordination strategy for dynamic target tracking in a contested undersea environment.
Integrate Llama 3.1 for intelligent mission re-planning and contextual threat assessment.
Deploy machine learning models using BentoML for edge processing on simulated UUVs.
Evaluate system robustness under communication constraints, environmental noise, and unexpected events.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The evaluation will focus on the system's ability to autonomously detect, classify, and track simulated undersea threats within a specified time frame, minimizing false positives and maximizing coverage, while leverag...
Challenge input
JSON configuration for UUV initial positions, sensor types (e.g., sonar range, FOV, noise), threat trajectory, and environmental noise parameters.
Agent execution
No agent tool is listed for this challenge.
Evaluated output
JSON log of UUV positions over time, detected threat positions, estimated threat classifications, and Llama 3.1's decision logs (e.g., re-planning...
- All UUVs must maintain simulated communication and demonstrate coordinated search patterns or tracking mane...
- The system must correctly classify the primary threat within a given confidence threshold (e.g., >80%) with...
- The anomaly detection model must be successfully deployed via BentoML and accessible to the simulated UUVs...
- Detection Latency target: 20
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- All UUVs must maintain simulated communication and demonstrate coordinated search patterns or tracking maneuvers.
- The system must correctly classify the primary threat within a given confidence threshold (e.g., >80%) within the specified time.
- The anomaly detection model must be successfully deployed via BentoML and accessible to the simulated UUVs for edge processing.
Metrics
- Detection Latency · Target 20
- Tracking Accuracy (RMS Error) · Target 1
- Coverage Efficiency · Target 95
- AI Decision Quality · Target 4
Workflow tasks
- Task 1
Input: JSON configuration for UUV initial positions, sensor types (e.g., sonar range, FOV, noise), threat trajectory, and environmental noise parameters.
Output: JSON log of UUV positions over time, detected threat positions, estimated threat classifications, and Llama 3.1's decision logs (e.g., re-planning commands, threat assessments).
Execution harness: Python
- Task 2
Input: JSON log from 'Threat Detection & Tracking Simulation' up to the event, plus event details (e.g., {"event_type": "new_threat", "location": [50, 50, -30], "time": 50}).
Output: JSON log of subsequent UUV positions, new threat detection status, and Llama 3.1's re-planning rationale and commands.
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.