---
title: >-
  Swiss Plasma Center at EPFL and Google DeepMind: Deep-reinforcement-learning magnetic
  control on the TCV tokamak
slug: deepmind-tcv-plasma-control
stable_id: 72fcb19d466f7070
company: Swiss Plasma Center at EPFL and Google DeepMind
function_code: engineering
pattern_codes:
  - autonomous_with_backstop
evidence_strength: verified
publication_tier: showcase
freshness: current
reviewed_at: '2026-08-23'
updated_at: '2026-08-22'
source_quality_summary: 1 peer reviewed; publication outcomes are verified.
caveat_summary: >-
  Research-machine demonstration, not commercial fusion-power operation. Demonstrated
  configurations do not establish generalization to other tokamaks.
collections:
  - embodied-work
  - negative-results
bundle_version: 1.0.0
bundle_fingerprint: sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93
canonical_url: https://brianletort.ai/transformations/deepmind-tcv-plasma-control
---

# Swiss Plasma Center at EPFL and Google DeepMind: Deep-reinforcement-learning magnetic control on the TCV tokamak

Operators set plasma objectives while a simulator-trained policy coordinates all coils on physical hardware.

Function: Engineering design. Patterns: Autonomous with a backstop. Evidence: verified.

Freshness: current. Reviewed: 2026-08-23. Updated: 2026-08-22.


Source quality: 1 peer reviewed; publication outcomes are verified.

## Before

1. **Control engineer** — Designs a controller for a predefined plasma state and configuration. (control: Physics models, engineering constraints, and machine protection.)
2. **Tokamak operator** — Runs the engineered controller during an experimental discharge. (control: Operational safety limits.)

## After

1. **Research operator** — Specifies plasma shape, current, and location objectives. (control: Simulator training plus physical and operational constraints.)
2. **RL controller** — Commands all magnetic coils at high frequency to track the requested plasma configuration. (control: TCV machine protection and operator-defined objectives.)

## Decision rights

The policy controls coils only within the approved discharge; researchers authorize experiments, objectives, constraints, and shutdowns.

## Exception path

Machine-protection systems and operators terminate or override a discharge when limits are breached.

## Outcomes

- **Plasma current and shape tracking error** (verified): A novel learned controller had not previously demonstrated direct full-coil magnetic control on tokamak hardware. → Across the full 0.1-1.0 second interval, RMSE was 0.62 kA for plasma current and 0.75 cm for shape.. Research-machine demonstration, not commercial fusion-power operation.
- **Configuration range** (verified): Engineering-driven control of pre-designed states. → The controller produced conventional, elongated, negative-triangularity, snowflake, and sustained two-droplet configurations.. Demonstrated configurations do not establish generalization to other tokamaks.

## Executive lesson

AI can move humans up from hand-crafting every control law to specifying objectives, but only when simulation, constraints, and machine protection form a hard envelope.

## Anti-pattern

Treating a research demonstration as evidence of autonomous commercial fusion operation.

## Questions for leaders

- How faithful is the simulator at edge conditions?
- Which shutdown rights remain independent of the model?
- What hardware transfer evidence is required?

## Sources

- [Magnetic control of tokamak plasmas through deep reinforcement learning](https://www.nature.com/articles/s41586-021-04301-9) — Nature