B2
Track B — Multi-Agent Orchestration & Decision Support

Environmental Monitoring Multi-Agent System

Systems where multiple specialized agents collaborate, hand off tasks, and reach a combined decision.

Background

Environmental monitoring usually happens in silos — air quality, water quality, and waste management are tracked by different teams with no combined risk picture for decision-makers.

Objective

Develop a multi-agent system where specialist agents each monitor one environmental signal, and a coordinator agent combines their findings into a single area risk report.

Key Features

1. Air quality monitoring agent 2. Water quality monitoring agent 3. Waste/litter detection agent (image-based) 4. Coordinator agent that combines all signals into one risk verdict 5. Dashboard showing individual and combined findings

Expected Solution

Three independent specialist agents, each processing one data type A coordinator agent that reasons across all three outputs A dashboard visualizing per-signal and combined risk

Suggested Technology Stack

Orchestration: LangGraph / CrewAI Vision model: YOLOv8 or MobileNet (transfer-learned) for litter detection Data: CPCB AQI API, Kaggle water-quality datasets, Kaggle/Roboflow litter datasets Frontend: Streamlit / React + Plotly

Expected Outcomes

A combined, reasoned environmental risk picture instead of siloed readings Faster identification of at-risk areas A reusable multi-agent coordination pattern for other monitoring domains

Possible Use Cases

Municipal bodies prioritizing cleanup/inspection efforts Citizens checking a combined environmental risk score for their area Researchers tracking multi-signal environmental trends

Evaluation Parameters

Independence and correctness of each specialist agent Quality of the coordinator's combined reasoning Dashboard clarity Genuine multi-agent hand-off (not just three features in one script)