RECORD
Robust Monocular Localization in Sparse HD Maps Leveraging Multi-Task Uncertainty Estimation
Robotics · Submitted 20 Oct 2021
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Extensive evaluation on the Lyft 5 dataset shows that, despite the sparsity of the map, our approach enables robust and accurate 6D localization in challenging urban scenarios
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Robust localization in dense urban scenarios using a low-cost sensor setup and sparse HD maps is highly relevant for the current advances in autonomous driving, but remains a challenging topic in research. We present a novel monocular localization approach based on a sliding-window pose graph that leverages predicted uncertainties for increased precision and robustness against challenging scenarios and per frame failures. To this end, we propose an efficient multi-task uncertainty-aware perception module, which covers semantic segmentation, as well as bounding box detection, to enable the localization of vehicles in sparse maps, containing only lane borders and traffic lights. Further, we design differentiable cost maps that are directly generated from the estimated uncertainties. This opens up the possibility to minimize the reprojection loss of amorphous map elements in an association free and uncertainty-aware manner. Extensive evaluation on the Lyft 5 dataset shows that, despite the sparsity of the map, our approach enables robust and accurate 6D localization in challenging urban scenarios
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Robust localization in dense urban scenarios using a low-cost sensor setup and sparse HD maps is highly relevant for the current advances in autonomous driving, but remains a challenging topic in research. | background 0.99 | 0.04 | 0.03 | asserted 1.00 |
| s1 | We present a novel monocular localization approach based on a sliding-window pose graph that leverages predicted uncertainties for increased precision and robustness against challenging scenarios and per frame failures. | contribution 0.88 | 0.90 | 0.16 | asserted 1.00 |
| s2 | To this end, we propose an efficient multi-task uncertainty-aware perception module, which covers semantic segmentation, as well as bounding box detection, to enable the localization of vehicles in sparse maps, containing only lane borders and traffic lights. | method 0.64 | 0.90 | 0.05 | asserted 0.97 |
| s3 | Further, we design differentiable cost maps that are directly generated from the estimated uncertainties. | method 0.81 | 0.94 | 0.03 | asserted 1.00 |
| s4 | This opens up the possibility to minimize the reprojection loss of amorphous map elements in an association free and uncertainty-aware manner. | interpretation 0.47 | 0.73 | 0.05 | hedged possibility 0.92 |
| s5 | Extensive evaluation on the Lyft 5 dataset shows that, despite the sparsity of the map, our approach enables robust and accurate 6D localization in challenging urban scenarios | result 1.00 | 0.96 | 0.89 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Robust localization in dense urban scenarios using a low-cost sensor setup and sparse HD maps is highly relevant for the current advances in autonomous driving, but remains a challenging topic in research. We present a novel monocular localization approach based on a sliding-window pose graph that leverages predicted uncertainties for increased precision and robustness against challenging scenarios and per frame failures. To this end, we propose an efficient multi-task uncertainty-aware perception module, which covers semantic segmentation, as well as bounding box detection, to enable the localization of vehicles in sparse maps, containing only lane borders and traffic lights. Further, we design differentiable cost maps that are directly generated from the estimated uncertainties. This opens up the possibility to minimize the reprojection loss of amorphous map elements in an association free and uncertainty-aware manner. Extensive evaluation on the Lyft 5 dataset shows that, despite the sparsity of the map, our approach enables robust and accurate 6D localization in challenging urban scenarios
Human review
How the literature uses it
2 arXiv papers cite this in 3 sentences
What it cites · 6 sentences without a confident label
Uses the method or tool of 3 papers
-
As proposed by [ref] , we utilize Normal-Inverse-Gamma (NIG) distribution as conjugate prior for evidential regression learning.
Deep Evidential Regression · Uncertainty-Aware Object Detection Head · 0.95
-
We utilize the sum of squares form of the loss $L(ζ)_i$ to penalize the misclassified pixels and the Kullback-Leibler (KL) divergence loss $ℒ^KL_i$ to predict high uncertainties for low evidence predictions for pixel $i$, as described by [ref] .
Evidential Deep Learning to Quantify Classification Uncertainty · Uncertainty-Aware Semantic Segmentation Head · 1.00
-
In addition, we show the calibration plot of the accuracy vs. the predicted probability, for semantic segmentation and predicted mean-variance (mVar) per bin vs. (mRMSE) per bin [ref] for bounding box detection, see fig:calibrationcurve.
Evaluating and Calibrating Uncertainty Prediction in Regression Tasks · Uncertainty Estimation · 0.97
Compares with 1 paper
-
The detector achieves an ENCE value of 11.5% which is reasonable in comparison to the value of 8.5% for [ref] reported by [ref] on the KITTI dataset.
Evaluating and Calibrating Uncertainty Prediction in Regression Tasks · Uncertainty Estimation · 0.97
Cites as background 2 papers
-
For example, methods such as [ref] , and [ref] employ modified versions of MC dropout to predict per pixel semantic and bounding box regression uncertainties, respectively.
Evaluating Bayesian Deep Learning Methods for Semantic Segmentation · Uncertainty Estimation · 0.81
-
Uncertainties are classified into aleatoric (data) uncertainty to quantify the noise in data and epistemic (model) uncertainty to quantify uncertainty in model prediction due to lack of training data or insufficient knowledge of the model [ref] .
What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? · Uncertainty Estimation · 1.00
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.