Enabling resilient autonomous motion planning requires robust predictions of surrounding road users' future behavior. In response to this need and the associated challenges, we introduce our model titled MTP-GO. The model encodes the scene using temporal graph neural networks to produce the inputs to an underlying motion model. The motion model is implemented using neural ordinary differential equations where the state-transition functions are learned with the rest of the model. Multimodal probabilistic predictions are obtained by combining the concept of mixture density networks and Kalman filtering. The results illustrate the predictive capabilities of the proposed model across various data sets, outperforming several state-of-the-art methods on a number of metrics.
Intelligent Autonomous Systems, Department of Computer Science, Technical University of Darmstadt(德意志技术大学达姆施塔特分校计算机科学系智能自主系统研究中心);Systems AI for Robot Learning, German Research Center for AI (DFKI)(德国人工智能研究中心(DFKI)机器人学习系统AI研究中心);Hessian Center for Artificial Intelligence (hessian.AI)(黑森人工智能中心(hessian.AI));Centre for Cognitive Science, Technical University of Darmstadt(德意志技术大学达姆施塔特分校认知科学中心);Automation & Control Institute, TU Wien(维也纳技术大学自动化与控制研究所);Austrian Institute of Technology (AIT), Vienna(奥地利技术研究所(AIT));VinRobotics and VinUniversity(VinRobotics和Vin大学)
CDFlow: Generative Gradient Flows for Configuration Space Distance Fields via Neural ODEs
Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ)(广东人工智能与数字经济实验室);College of Design and Engineering(设计与工程学院);College of Computer Science and Software Engineering(计算机科学与软件工程学院);School of Information Technology(信息技术学院)
A Legendre-Gauss Pseudospectral Collocation Method for Trajectory Optimization in Second Order Systems
Institut de Robòtica i Informàtica Industrial, CSIC-UPC(西班牙国家研究委员会-加泰罗尼亚理工大学工业机器人与信息技术研究所)
伪谱配点法是求解最优控制问题的有力工具。虽然这些方法通常假设动力学以一阶形式$\dot{x} = f (x, u, t)$给出,其中x是状态,u是控制向量,但机器人系统通常由$\ddot{q} = g(q, \dot{q}, u, t)$形式的二阶ODE控制,其中q是配置。为了将二阶常微分方程转化为一阶常微分方程,通常的方法是引入一个速度变量v,并使它与q的时间导数重合。Lobatto方法通过构造来赋予这个约束,因为它们描述v的轨迹的多项式是q的多项式的时间导数,但高斯和Radau方法不能这么说。这对于这样的方法是有问题的,因为它们不能保证$\ddot{q} = g(q, \dot{q}, u, t)$在配置点处。在其消极的一面,Lobatto方法不能用于解决初始值问题,因为给定配置点处的u值,它们会生成状态的过约束方程组。在本文中,我们提出了一个Legendre-Gauss配置方法,保留了通常的Lobatto,Gauss和Radau方法的优点,同时避免了它们的缺点。我们提出的配置格式适用于求解初值问题,保持了v和q的多项式之间的一致性,并保证了配置点处的$\ddot{q} = g(q, \dot{q}, u, t)$。
Continuous-Time Gaussian Belief Trees for Motion Planning
Department of Aerospace Engineering Sciences at the University of Colorado Boulder(科罗拉多大学博尔德分校航空航天工程科学系);Department of Aerospace and Ocean Engineering at Virginia Tech(弗吉尼亚理工学院航空航天与海洋工程系)