DiffCo: Autodifferentiable Proxy Collision Detection With Multiclass Labels for Safety-Aware Trajectory Optimization
DiffCo: Autodifferentiable Proxy Collision Detection With Multiclass Labels for Safety-Aware Trajectory Optimization
复制标题
DiffCo：具有多类标签的自动可微代理碰撞检测，用于安全感知轨迹优化
DOI：
10.1109/tro.2022.3153789
复制
发表时间：
2021
影响因子：
7.8
通讯作者：
Michael C. Yip
中科院分区：
文献类型：
--
作者：
Yuheng Zhi;Nikhil Das;Michael C. Yip
轨迹优化算法的目标是实现起始状态和目标状态之间的最优无碰撞路径。在现实世界的场景中，环境可能是复杂和不均匀的，机器人需要能够衡量一个状态是否会与各种物体发生碰撞，以满足一些安全指标。碰撞检测器应该在计算上高效，并且理想情况下，分析可微，以促进优化期间稳定和快速的梯度下降。然而，今天的方法缺乏一个优雅的方法来检测碰撞微分，而依赖于数值梯度，可能是不稳定的。我们提出了DiffCo，第一个完全自动微分的非参数碰撞检测模型。它的非参数行为允许人们在飞行中计算碰撞边界并更新它们，不需要预先训练，并允许它在动态环境中不断更新。它通过反向传播为轨迹优化提供了强大的梯度，并且通常比其几何对应物的计算速度快10-100倍。DiffCo还扩展到对不同对象碰撞类进行建模，以进行语义上的轨迹优化。
The objective of trajectory optimization algorithms is to achieve an optimal collision-free path between start and goal states. In real-world scenarios, where environments can be complex and nonhomogeneous, a robot needs to be able to gauge whether a state will be in collision with various objects in order to meet some safety metrics. The collision detector should be computationally efficient and, ideally, analytically differentiable to facilitate stable and rapid gradient descent during optimization. However, methods today lack an elegant approach to detect collision differentiably, relying rather on numerical gradients that can be unstable. We present DiffCo, the first, fully autodifferentiable, nonparametric model for collision detection. Its nonparametric behavior allows one to compute collision boundaries on the fly and update them, requiring no pretraining and allowing it to update continuously in dynamic environments. It provides robust gradients for trajectory optimization via backpropagation and is often 10–100 times faster to compute than its geometric counterparts. DiffCo also extends trivially to modeling different object collision classes for semantically informed trajectory optimization.
影响因子：
5.2
作者：
B. Wilcox;Michael C. Yip
通讯作者：
B. Wilcox;Michael C. Yip