A scalable multi-agent deep reinforcement learning in thermoforming: An experimental evaluation of thermal control by infrared camera-based feedback
Bibliographic record
Abstract
This manuscript presents the development of multi-agent Deep Reinforcement Learning (DRL) for radiation thermal control in thermoforming processes involving multiple heaters. The complexity of such control systems is characterized by significant action and state spaces, where the actions of all actuators collectively influence the system's output. This complexity introduces substantial challenges regarding the computational demands for offline training of learning-based algorithms and the online computational costs associated with a real-world controller deployment. The study presents a novel approach to training an adaptive and robust DRL agent system that can control a single heating element on the thermoplastic sheet while dynamically considering interactive effects from nearby heaters. Results demonstrated that upon deploying the pre-trained agent for each heater within the heater bank, the group of agents could then regulate the temperature of the sheet to any physically feasible output temperature profile. In contrast to the conventional DRL approach, where a single agent manages all heaters, the multi-agent DRL method boasted that an offline training process was 110 times faster, coupled with an 8 times reduction in the final error margin on the simulator. The experimental data, conducted on a laboratory-scale setup, confirmed the performance of the proposed model, with a final absolute error under 4 ° C . Regardless of the number of heaters, the multi-agent DRL approach exhibited accurate and robust performance. Its advantage was that it incurred no significant offline and online computational burden when the number of heating elements increased, deemed a promising notion for industrial-scale applications.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".