arXiv subjects

Christian Bohn

Publications and source records attributed to Christian Bohn.

6 recordsLinked to original sources

arXiv2026

Rectify, Don't Regret: Avoiding Pitfalls of Differentiable Simulation in Trajectory Prediction

Current open-loop trajectory models struggle in real-world autonomous driving because minor initial deviations often cascade into compounding errors, pushing the agent into out-of-distribution states. While fully differentiable closed-loop simulators attempt to address this, they suffer from shortcut learning: the loss gradients flow backward through induced state inputs, inadvertently leaking future ground truth information directly into the model's own previous predictions. The model exploits these signals to artificially avoid drift, non-causally "regretting" past mistakes rather than learning genuinely reactive recovery. To address this, we introduce a detached receding horizon rollout. By explicitly severing the computation graph between simulation steps, the model learns genuine recovery behaviors from drifted states, forcing it to "rectify" mistakes rather than non-causally optimizing past predictions. Extensive evaluations on the nuScenes and DeepScenario datasets show our approach yields more robust recovery strategies, reducing target collisions by up to 33.24% compared to fully differentiable closed-loop training at high replanning frequencies. Furthermore, compared to standard open-loop baselines, our non-differentiable framework decreases collisions by up to 27.74% in dense environments while simultaneously improving multi-modal prediction diversity and lane alignment.

cs.RO↗

arXiv2026

Faster Training, Fewer Labels: Self-Supervised Pretraining for Fine-Grained BEV Segmentation

Dense Bird's Eye View (BEV) semantic maps are central to autonomous driving, yet current multi-camera methods depend on costly, inconsistently annotated BEV ground truth. We address this limitation with a two-phase training strategy for fine-grained road marking segmentation that removes full supervision during pretraining and halves the amount of training data during fine-tuning while still outperforming the comparable supervised baseline model. During the self-supervised pretraining, BEVFormer predictions are differentiably reprojected into the image plane and trained against multi-view semantic pseudo-labels generated by the widely used semantic segmentation model Mask2Former. A temporal loss encourages consistency across frames. The subsequent supervised fine-tuning phase requires only 50% of the dataset and significantly less training time. With our method, the fine-tuning benefits from rich priors learned during pretraining boosting the performance and BEV segmentation quality (up to +2.5pp mIoU over the fully supervised baseline) on nuScenes. It simultaneously halves the usage of annotation data and reduces total training time by up to two thirds. The results demonstrate that differentiable reprojection plus camera perspective pseudo labels yields transferable BEV features and a scalable path toward reduced-label autonomous perception.

cs.CV↗

arXiv2025

Efficient Inter-Task Attention for Multitask Transformer Models

In both Computer Vision and the wider Deep Learning field, the Transformer architecture is well-established as state-of-the-art for many applications. For Multitask Learning, however, where there may be many more queries necessary compared to single-task models, its Multi-Head-Attention often approaches the limits of what is computationally feasible considering practical hardware limitations. This is due to the fact that the size of the attention matrix scales quadratically with the number of tasks (assuming roughly equal numbers of queries for all tasks). As a solution, we propose our novel Deformable Inter-Task Self-Attention for Multitask models that enables the much more efficient aggregation of information across the feature maps from different tasks. In our experiments on the NYUD-v2 and PASCAL-Context datasets, we demonstrate an order-of-magnitude reduction in both FLOPs count and inference latency. At the same time, we also achieve substantial improvements by up to 7.4% in the individual tasks' prediction quality metrics.

cs.CV↗

arXiv2024

Task Weighting through Gradient Projection for Multitask Learning

In multitask learning, conflicts between task gradients are a frequent issue degrading a model's training performance. This is commonly addressed by using the Gradient Projection algorithm PCGrad that often leads to faster convergence and improved performance metrics. In this work, we present a method to adapt this algorithm to simultaneously also perform task prioritization. Our approach differs from traditional task weighting performed by scaling task losses in that our weighting scheme applies only in cases where tasks are in conflict, but lets the training proceed unhindered otherwise. We replace task weighting factors by a probability distribution that determines which task gradients get projected in conflict cases. Our experiments on the nuScenes, CIFAR-100, and CelebA datasets confirm that our approach is a practical method for task weighting. Paired with multiple different task weighting schemes, we observe a significant improvement in the performance metrics of most tasks compared to Gradient Projection with uniform projection probabilities.

cs.LG↗

arXiv2021

Blind Identification of State-Space Models in Physical Coordinates

Blind identification is popular for modeling a system without the input information, such as in the research areas of structural health monitoring and audio signal processing. Existing blind identification methods have both advantages and disadvantages, in this paper, we briefly outline current methods and propose a novel blind identification method for identifying state-space models in physical coordinates. The idea behind this proposed method is first to regard the collected input data of a state-space model as a part of a periodic signal sequence, and then transform the state-space model with input and output into a model without input by augmenting the state-space model with the input model (which is a periodic signal model), and afterwards use merely the output information to identify a state-space model up to a similarity transformation, and finally derive the state-space model in physical coordinates by using a unique similarity transformation. With the above idea, physical parameters and modal parameters of a state-space system can be obtained. Both numerical and practical examples were used to validate the proposed method. The result showed the effectiveness of the novel blind identification method.

eess.SY↗

arXiv2016

Parameter Optimisation of a Virtual Synchronous Machine in a Microgrid

Parameters of a virtual synchronous machine in a small microgrid are optimised. The dynamical behaviour of the system is simulated after a perturbation, where the system needs to return to its steady state. The cost functional evaluates the system behaviour for different parameters. This functional is minimised by Parallel Tempering. Two perturbation scenarios are investigated and the resulting optimal parameters agree with analytical predictions. Dependent on the focus of the optimisation different optima are obtained for each perturbation scenario. During the transient the system leaves the allowed voltage and frequency bands only for a short time if the perturbation is within a certain range.

math.OC↗