Institution profile

TNO

Academic institutioneurope · nl
Official website
Research library55linked papers
Opportunities0open roles
Selected work

Representative Papers

An AI-assisted conditioning and geological interpretation workflow for usage in implicit geological modeling

Oct 07, 2026

This study addresses the challenges of low efficiency, significant subjective bias, and heavy reliance on manual labor in seismic interpretation for implicit geological modeling by proposing an AI-assisted interpretation toolkit that requires minimal labeled data. Methodologically, self-supervised and semi-supervised contrastive learning convolutional neural networks are employed to achieve signal enhancement, noise suppression, and data interpolation, enabling the automated extraction of horizons and faults. This workflow substantially improves both the modeling speed and reproducibility of shallow-to-deep seismic data. The successful application to the top boundary of the Maassluis Formation in the Netherlands validates the maturity and practical value of the proposed approach for real-world geological decision-making.

0 citationsRead paper

LiFT: Loop Flow Transformers

Oct 04, 2026

This study addresses the high computational costs incurred when diffusion models rely on increasing parameters or complex architectures to enhance performance. To this end, we propose the Recurrent Flow Transformer, which iteratively applies a shared DiT core and introduces a novel single-step regression training scheme based on continuous depth indexing. This design enables test-time compute scaling without retraining, supporting both early exit and infinite-loop inference beyond the trained depth. Evaluated on ImageNet 256×256, our method reduces FID by 3.34 points while decreasing parameter count by 60% and lowering training and inference FLOPs by 32% and 52%, respectively. These results demonstrate an efficient and scalable approach to generative modeling.

0 citationsRead paper

Towards Safer Autonomous Driving in an Open World: A Dual-Process Approach

Oct 02, 2026

This study addresses the safety and regulatory compliance deficiencies of autonomous driving in out-of-distribution scenarios by proposing a planning framework grounded in dual-process theory. The framework integrates intuitive neural network-based planning with deliberative Model Predictive Control (MPC), incorporating a metacognitive mechanism and knowledge graphs to quantify contextual risk fields. This design enables dynamic switching between the two modes to handle unfamiliar road conditions. Experimental evaluations in the CARLA simulator demonstrate that the proposed approach reduces collision rates by 89% while significantly improving compliance with complex right-of-way regulations. By effectively balancing computational efficiency and operational safety, this work establishes a novel paradigm for autonomous driving in open-world environments.

0 citationsRead paper
Recent publications

Latest Papers

An AI-assisted conditioning and geological interpretation workflow for usage in implicit geological modeling

Oct 07, 2026

This study addresses the challenges of low efficiency, significant subjective bias, and heavy reliance on manual labor in seismic interpretation for implicit geological modeling by proposing an AI-assisted interpretation toolkit that requires minimal labeled data. Methodologically, self-supervised and semi-supervised contrastive learning convolutional neural networks are employed to achieve signal enhancement, noise suppression, and data interpolation, enabling the automated extraction of horizons and faults. This workflow substantially improves both the modeling speed and reproducibility of shallow-to-deep seismic data. The successful application to the top boundary of the Maassluis Formation in the Netherlands validates the maturity and practical value of the proposed approach for real-world geological decision-making.

0 citationsRead paper

LiFT: Loop Flow Transformers

Oct 04, 2026

This study addresses the high computational costs incurred when diffusion models rely on increasing parameters or complex architectures to enhance performance. To this end, we propose the Recurrent Flow Transformer, which iteratively applies a shared DiT core and introduces a novel single-step regression training scheme based on continuous depth indexing. This design enables test-time compute scaling without retraining, supporting both early exit and infinite-loop inference beyond the trained depth. Evaluated on ImageNet 256×256, our method reduces FID by 3.34 points while decreasing parameter count by 60% and lowering training and inference FLOPs by 32% and 52%, respectively. These results demonstrate an efficient and scalable approach to generative modeling.

0 citationsRead paper

Towards Safer Autonomous Driving in an Open World: A Dual-Process Approach

Oct 02, 2026

This study addresses the safety and regulatory compliance deficiencies of autonomous driving in out-of-distribution scenarios by proposing a planning framework grounded in dual-process theory. The framework integrates intuitive neural network-based planning with deliberative Model Predictive Control (MPC), incorporating a metacognitive mechanism and knowledge graphs to quantify contextual risk fields. This design enables dynamic switching between the two modes to handle unfamiliar road conditions. Experimental evaluations in the CARLA simulator demonstrate that the proposed approach reduces collision rates by 89% while significantly improving compliance with complex right-of-way regulations. By effectively balancing computational efficiency and operational safety, this work establishes a novel paradigm for autonomous driving in open-world environments.

0 citationsRead paper