Software-Defined PLC Ladder Logic Generator with Claude Agents SDK and H2O AutoML
Translating natural language automation descriptions into standardized IEC 61131-3 PLC Structured Text (ST) remains prone to syntax and logic errors. You will construct a software-defined controls agent using Claude Agents SDK and H2O AutoML to analyze sensor telemetry patterns and synthesize deterministic IEC 61131-3 ST code. The generated code must compile clean without syntax errors and satisfy state machine requirements.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an IEC 61131-3 PLC code generation engine using Claude Agents SDK with H2O AutoML predictive dynamic optimization.
How work is evaluated
Validates generated IEC 61131-3 code by parsing and compiling generated Structured Text routines.
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.
plc_syntax_check
Verifies program structure adheres to IEC 61131-3 keywords (PROGRAM, VAR_INPUT, END_PROGRAM).
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
code_compilation_success_rate
Percentage of generated ST scripts passing syntax parsing • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement Claude Agents SDK workflows featuring extended thinking for complex PLC control loops
Integrate H2O AutoML models to predict optimal loop tuning constants (Kp, Ki, Kd) from process telemetry
Generate syntactically valid IEC 61131-3 Structured Text (ST) code blocks from process state specifications
Verify control logic compilation against open-source IEC compilers (e.g., matiec / OpenPLC)
Reference links and supporting material
Process descriptions and telemetry step-response curves for industrial mixing tanks and continuous heat exchangers.
Challenge setup
Validates generated IEC 61131-3 code by parsing and compiling generated Structured Text routines.
Challenge input
JSON specifying input sensors, output actuators, state transitions, and H2O PID parameter estimates
Claude Agents SDK
Required framework assignment.
AutoML (H2O)
Required tool assignment.
Evaluated output
JSON with structured_text_code string and compilation_status boolean
- Verifies program structure adheres to IEC 61131-3 keywords (PROGRAM, VAR_INPUT, END_PROGRAM).
- Percentage of generated ST scripts passing syntax parsing • target: 1 • range: 0-1
- Benchmark: PLC Code Synthesis Benchmark
- Code Compilation Success Rate target: 100%
- 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
- Claude Code · Required
- Claude Code · Required
- Claude 4.1 Opus · Optional
- Claude 4 Sonnet · Optional
Evaluation contract
- plc_syntax_check · Weight 1
- code_compilation_success_rate · Weight 1
Metrics
- code_compilation_success_rate · Target 1
Benchmarks
- PLC Code Synthesis Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON specifying input sensors, output actuators, state transitions, and H2O PID parameter estimates
Output: JSON with structured_text_code string and compilation_status boolean
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent Orchestration Framework
- AutoML (H2O) · ML Control Parameter Predictor
Recipe freeze marker
No recipe freeze time is recorded.