Pedestrian motion prediction is a fundamental task for autonomous robots and vehicles to operate safely. In recent years many complex approaches based on neural networks have been proposed address this problem. work we show that - surprisingly simple Constant Velocity Model can outperform even state-of-the-art models. This indicates either are not able make use of the additional information they provided with, or as relevant commonly believed. Therefore, analyze how process their input it...

The environmental perception of an autonomous vehicle is limited by its physical sensor ranges and algorithmic performance, as well occlusions that degrade understanding ongoing traffic situation. This not only poses a significant threat to safety limits driving speeds, but it can also lead inconvenient maneuvers. Intelligent Infrastructure Systems help alleviate these problems. An System fill in the gaps vehicle’s extend field view providing additional detailed information about...

Most intelligent transportation systems use a combination of radar sensors and cameras for robust vehicle perception. The calibration these heterogeneous sensor types in an automatic fashion during system operation is challenging due to differing physical measurement principles the high sparsity traffic radars. We propose - best our knowledge first data-driven method rotational radar-camera without dedicated targets. Our approach based on coarse fine convolutional neural network. employ...

The future motion of traffic participants is inherently uncertain. To plan safely, therefore, an autonomous agent must take into account multiple possible trajectory outcomes and prioritize them. Recently, this problem has been addressed with generative neural networks. However, most models either do not learn the true underlying distribution reliably, or allow predictions to be associated likelihoods. In our work, we model prediction directly as a density estimation normalizing flow between...

Ensuring the safety of autonomous vehicles remains challenging given uncertainty in sensing other road users. Moreover, separate specifications for perception and planning components impede assessing overall system safety. This work provides a probabilistic approach to calculate envelopes under uncertainty. The envelope definition is based on risk threshold. It limits cumulative probability that actual fully observable environment physically more extensive than applied solved using iterative...

The environmental perception of an autonomous vehicle is limited by its physical sensor ranges and algorithmic performance, as well occlusions that degrade understanding ongoing traffic situation. This not only poses a significant threat to safety limits driving speeds, but it can also lead inconvenient maneuvers. Intelligent Infrastructure Systems help alleviate these problems. An System fill in the gaps vehicle's extend field view providing additional detailed information about...

In this paper, we provide the first steps towards real-time, large-scale prediction of lifetime information diffusion processes.

Most intelligent transportation systems use a combination of radar sensors and cameras for robust vehicle perception. The calibration these heterogeneous sensor types in an automatic fashion during system operation is challenging due to differing physical measurement principles the high sparsity traffic radars. We propose - best our knowledge first data-driven method rotational radar-camera without dedicated targets. Our approach based on coarse fine convolutional neural network. employ...

Pedestrian motion prediction is a fundamental task for autonomous robots and vehicles to operate safely. In recent years many complex approaches based on neural networks have been proposed address this problem. work we show that - surprisingly simple Constant Velocity Model can outperform even state-of-the-art models. This indicates either are not able make use of the additional information they provided with, or as relevant commonly believed. Therefore, analyze how process their input it...

Ensuring the safety of autonomous vehicles, given uncertainty in sensing other road users, is an open problem. Moreover, separate specifications for perception and planning components raise how to assess overall system safety. This work provides a probabilistic approach calculate envelopes under uncertainty. The envelope definition based on risk threshold. It limits cumulative probability that actual fully observable environment larger than applied solved using iterative worst-case analysis...

The future motion of traffic participants is inherently uncertain. To plan safely, therefore, an autonomous agent must take into account multiple possible trajectory outcomes and prioritize them. Recently, this problem has been addressed with generative neural networks. However, most models either do not learn the true underlying distribution reliably, or allow predictions to be associated likelihoods. In our work, we model prediction directly as a density estimation normalizing flow between...