arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2101.05893cs.ROcs.CV
收藏

多Agent环境下的实例感知预测导航

Instance-Aware Predictive Navigation in Multi-Agent Environments

Carnegie Mellon University(卡内基梅隆大学);University of California, Berkeley(加州大学伯克利分校);ETH Zurich(苏黎世联邦理工学院)

Jinkun Cao, Xin Wang, Trevor Darrell, Fisher Yu

摘要:

在这项工作中,我们的目标是实现高效的端到端学习的动态多智能体环境中的驾驶策略。在对象级别预测和预测未来事件对于做出明智的驾驶决策至关重要。我们提出了一个实例感知预测控制(IPC)的方法,预测代理之间的相互作用以及未来的场景结构。我们采用了一种新的多实例事件预测模块来估计可能的互动代理在自我为中心的观点,条件是所选择的动作序列的自我车辆。为了决定每一步的动作,我们通过重复采样可能的动作序列,根据预测模块的输出来寻找可以导致安全未来状态的动作序列。我们设计了一个顺序动作采样策略,以更好地利用场景级和实例级的预测状态。我们的方法建立了一个新的艺术状态在具有挑战性的CARLA多智能体驾驶模拟环境中没有专家演示,提供更好的解释性和样本效率。

英文摘要:

In this work, we aim to achieve efficient end-to-end learning of driving policies in dynamic multi-agent environments. Predicting and anticipating future events at the object level are critical for making informed driving decisions. We propose an Instance-Aware Predictive Control (IPC) approach, which forecasts interactions between agents as well as future scene structures. We adopt a novel multi-instance event prediction module to estimate the possible interaction among agents in the ego-centric view, conditioned on the selected action sequence of the ego-vehicle. To decide the action at each step, we seek the action sequence that can lead to safe future states based on the prediction module outputs by repeatedly sampling likely action sequences. We design a sequential action sampling strategy to better leverage predicted states on both scene-level and instance-level. Our method establishes a new state of the art in the challenging CARLA multi-agent driving simulation environments without expert demonstration, giving better explainability and sample efficiency.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑