Multi-Agent Protocol for Smart City Robotaxi Coordination
This challenge focuses on designing a multi-agent system for smart city traffic management. You will build an A2A protocol-enabled system using Langroid agents powered by DeepSeek to coordinate different robotaxi fleets. The system must optimize traffic flow, dynamically re-route vehicles, and resolve conflicts by leveraging real-time traffic data, extended thinking for predictive routing, and few-shot learning to adapt to evolving urban conditions. Emphasis will be placed on seamless inter-fleet cooperation and robust decision-making in a complex, dynamic environment. This project requires implementing sophisticated agent communication and autonomous reasoning. You will integrate MCP tools for real-time urban data streams and orchestrate large-scale agent deployments using OpenAI Swarm, pushing the envelope for urban mobility solutions.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an A2A protocol multi-agent system with Langroid and DeepSeek V3 for smart city robotaxi coordination, using extended thinking and few-shot learning for adaptive traffic management and OpenAI Swarm for scaling.
How work is evaluated
The evaluation will focus on the multi-agent system's ability to optimize traffic flow, reduce congestion, and resolve routing conflicts in a simulated urban environment, emphasizing inter-fleet A2A communication, adaptive reasoning, and resilience.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Design and implement an A2A protocol for cross-fleet agent communication.
Develop Langroid agents for real-time traffic analysis and routing decisions using DeepSeek V3.
Orchestrate and scale agents using OpenAI Swarm for urban environments.
Integrate MCP tools for real-time urban data streams.
Apply extended thinking and few-shot learning for adaptive traffic management.
Reference links and supporting material
Sample data for 1 tasks
Challenge setup
The evaluation will focus on the multi-agent system's ability to optimize traffic flow, reduce congestion, and resolve routing conflicts in a simulated urban environment, emphasizing inter-fleet A2A communication, ada...
Challenge input
{"traffic_state": {"road_closures": ["string", ...], "event_locations": ["string", ...], "current_congestion": {"street_name": "level"}, "new_patte...
Agent execution
No agent tool is listed for this challenge.
Evaluated output
{"fleet_A_new_routes": [{"id": "uuid", "optimized_path": ["node", ...]}], "fleet_B_new_routes": [{"id": "uuid", "optimized_path": ["node", ...]}],...
- Optimized routes must not create new, severe congestion points or cause gridlock in the simulated environment.
- Communication logs must show successful and structured A2A protocol interactions between fleet agents and a...
- The DeepSeek V3 agent's reasoning should demonstrate successful application of few-shot learning to adapt t...
- Overall Travel Time Reduction target: 0.25
- 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
- Optimized routes must not create new, severe congestion points or cause gridlock in the simulated environment.
- Communication logs must show successful and structured A2A protocol interactions between fleet agents and a central coordinator.
- The DeepSeek V3 agent's reasoning should demonstrate successful application of few-shot learning to adapt to specific, new traffic pattern examples provided in the input.
Metrics
- Overall Travel Time Reduction · Target 0.25
- Congestion Index Reduction · Target 0.2
- Decision-Making Latency (ms) · Target 1500
Workflow tasks
- Task 1
Input: {"traffic_state": {"road_closures": ["string", ...], "event_locations": ["string", ...], "current_congestion": {"street_name": "level"}, "new_patterns_examples": [{"input": "scenario_desc", "output": "expected_agent_action"}, ...]}, "fleet_A_trips": [{"start": "loc", "end": "loc", "id": "uuid"}], "fleet_B_trips": [{"start": "loc", "end": "loc", "id": "uuid"}]}
Output: {"fleet_A_new_routes": [{"id": "uuid", "optimized_path": ["node", ...]}], "fleet_B_new_routes": [{"id": "uuid", "optimized_path": ["node", ...]}], "a2a_communication_logs": [{"sender": "string", "receiver": "string", "message": "string", "timestamp": "datetime"}], "deepseek_reasoning_logs": "string"}
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.