Mutual localization provides a consensus of reference frame as an essential basis for cooperation in multirobot systems. Previous works have developed certifiable and robust solvers for relative transformation estimation between each pair of robots. However, recovering relative poses for robotic swarm with partially mutual observations is still an unexploited problem. In this paper, we present a complete algorithm for it with optimality, scalability and robustness. Firstly, we fuse all odometry and bearing measurements in a unified minimization problem among the Stiefel manifold. Furthermore, we relax the original non-convex problem into a semi-definite programming (SDP) problem with a strict tightness guarantee. Then, to hold the exactness in noised cases, we add a convex (linear) rank cost and apply a convex iteration algorithm. We compare our approach with local optimization methods on extensive simulations with different robot amounts under various noise levels to show our global optimality and scalability advantage. Finally, we conduct real-world experiments to show the practicality and robustness.
翻译:作为多机器人系统合作的必要基础,相互定位提供了一个一致的参照框架。 先前的工程已经开发了每对机器人相对转换估计的可验证和强大的解决方案。 然而,通过部分相互观测来恢复机器人群的相对构成仍然是一个尚未开发的问题。 在本文中,我们提出了一个完整的机器人群算法,其优化性、可缩放性和坚固性。 首先,我们将所有odo测量法和测量方法结合到Stiefel系统之间的统一最小化问题中。 此外,我们用严格的紧凑性保证,将原非convex问题放松到半确定性程序(SDP)问题中。 然后,为了在编定的案例中保持准确性,我们添加一个线性级成本,并应用一个连接性轴轴轴转换算法。 我们将我们的方法与广泛模拟的本地优化方法作比较,在不同噪音级别下的不同机器人数量,以显示我们的全球最佳性和可缩放优势。 最后,我们进行现实世界实验,以显示实际性和稳健度。