Remote-Controlled RTG Crane Scheduling Agent with AutoGen
Mitsui O.S.K. Lines (MOL) is expanding the Port of Kobe terminal by deploying remote-controlled rubber-tired gantry cranes (RTGs). Autonomous scheduling agents must bridge human remote operators and automated crane move instructions to maximize container moves per hour while eliminating operator idle time. In this challenge, you will implement an AutoGen multi-agent system consisting of an Automated Job Dispatcher, a Remote Operator Load Balancer, and a Safety Supervision Agent. The agents collaborate to assign crane pickup jobs based on operator shift availability and yard congestion.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an AutoGen multi-agent crane dispatch framework for remote-controlled RTGs at container terminals.
How work is evaluated
Evaluates move dispatch efficiency and adherence to operator fatigue rules.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
fatigue_safety_rule
Operators with fatigue > 0.8 must not be assigned remote control moves.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
moves_per_hour
Simulated crane moves completed per hour • target: 32 • range: 0-60
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Set up an AutoGen GroupChat with specialized job dispatcher and safety supervisor agents
Model remote-control desk operator fatigue and shift assignment constraints
Optimize container pick-and-place sequence using discrete move scheduling tools
Implement emergency human-in-the-loop handoff triggers for safety anomalies
Reference links and supporting material
Port of Kobe crane move logs, remote desk telemetry, and operator workload shift metrics.
Challenge setup
Evaluates move dispatch efficiency and adherence to operator fatigue rules.
Challenge input
JSON container move queue, operator fatigue scores, and yard block locations
AutoGen
Required framework for collaborative agent decision-making.
Evaluated output
JSON schedule plan with operator-to-crane assignments and safety approval boolean
- Operators with fatigue > 0.8 must not be assigned remote control moves.
- Simulated crane moves completed per hour • target: 32 • range: 0-60
- Benchmark: RoboOps-Bench
- Moves Per Hour target: 32
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
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
- AutoGen · Required
- Shakudo · Optional
- FLAML · Optional
Evaluation contract
- fatigue_safety_rule · Weight 1
- moves_per_hour · Weight 1
Metrics
- moves_per_hour · Target 32
Benchmarks
- RoboOps-Bench
- robotics
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON container move queue, operator fatigue scores, and yard block locations
Output: JSON schedule plan with operator-to-crane assignments and safety approval boolean
Execution harness: Python
Evaluation tools
- AutoGen · Multi-Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.