This paper presents a white-box intention-aware decision-making for the
handling of interactions between a pedestrian and an automated vehicle (AV) in
an unsignalized street crossing scenario. Moreover, a design framework has been
developed, which enables automated parameterization of the decision-making.
This decision-making is designed in such a manner that it can understand
pedestrians in urban traffic and can react accordingly to their intentions.
That way, a human-like response to the actions of the pedestrian is ensured,
leading to a higher acceptance of AVs. The core notion of this paper is that
the intention prediction of the pedestrian to cross the street and
decision-making are divided into two subsystems. On the one hand, the intention
detection is a data-driven, black-box model. Thus, it can model the complex
behavior of the pedestrians. On the other hand, the decision-making is a
white-box model to ensure traceability and to enable a rapid verification and
validation of AVs. This white-box decision-making provides human-like behavior
and a guaranteed prevention of deadlocks. An additional benefit is that the
proposed decision-making requires low computational resources only enabling
real world usage. The automated parameterization uses a particle swarm
optimization and compares two different models of the pedestrian: The social
force model and the Markov decision process model. Consequently, a rapid design
of the decision-making is possible and different pedestrian behaviors can be
taken into account. The results reinforce the applicability of the proposed
intention-aware decision-making.

本文提出了一种针对行人和自动驾驶汽车在无红绿灯的过街场景中交互的白盒意图感知决策方法，并提供了一个设计框架。作者采用黑盒意图检测和白盒决策的混合方法实现了类人行为和预防死锁，此方法计算量低并能考虑不同的行人行为。

混合交叉口场景下的意图感知决策

Intention-Aware Decision-Making for Mixed Intersection Scenarios

Understanding and predicting the intention of pedestrians is essential to
enable autonomous vehicles and mobile robots to navigate crowds. This problem
becomes increasingly complex when we consider the uncertainty and multimodality
of pedestrian motion, as well as the implicit interactions between members of a
crowd, including any response to a vehicle. Our approach, Probabilistic Crowd
GAN, extends recent work in trajectory prediction, combining Recurrent Neural
Networks (RNNs) with Mixture Density Networks (MDNs) to output probabilistic
multimodal predictions, from which likely modal paths are found and used for
adversarial training. We also propose the use of Graph Vehicle-Pedestrian
Attention Network (GVAT), which models social interactions and allows input of
a shared vehicle feature, showing that inclusion of this module leads to
improved trajectory prediction both with and without the presence of a vehicle.
Through evaluation on various datasets, we demonstrate improvements on the
existing state of the art methods for trajectory prediction and illustrate how
the true multimodal and uncertain nature of crowd interactions can be directly
modelled.

本文介绍了一种使用概率多模式预测方法和社交影响建模的新型 GAN，可有效预测人群行为和行人意图，以达到更好的自动驾驶和移动机器人导航效果。