arXiv subjects

Xiangyu Kong

Publications and source records attributed to Xiangyu Kong.

45 records · Page 3Linked to original sources

arXiv2020

Pose-Assisted Multi-Camera Collaboration for Active Object Tracking

Active Object Tracking (AOT) is crucial to many visionbased applications, e.g., mobile robot, intelligent surveillance. However, there are a number of challenges when deploying active tracking in complex scenarios, e.g., target is frequently occluded by obstacles. In this paper, we extend the single-camera AOT to a multi-camera setting, where cameras tracking a target in a collaborative fashion. To achieve effective collaboration among cameras, we propose a novel Pose-Assisted Multi-Camera Collaboration System, which enables a camera to cooperate with the others by sharing camera poses for active object tracking. In the system, each camera is equipped with two controllers and a switcher: The vision-based controller tracks targets based on observed images. The pose-based controller moves the camera in accordance to the poses of the other cameras. At each step, the switcher decides which action to take from the two controllers according to the visibility of the target. The experimental results demonstrate that our system outperforms all the baselines and is capable of generalizing to unseen environments. The code and demo videos are available on our website https://sites.google.com/view/pose-assistedcollaboration.

cs.CV↗

arXiv2019

Experimental observation of information flow in the anti-$\mathcal{PT}$-symmetric system

The recently theoretical and experimental researches related to $\mathcal{PT}$-symmetric system have attracted unprecedented attention because of various novel features and potentials in extending canonical quantum mechanics. However, as the counterpart of $\mathcal{PT}$-symmetry, there are only a few researches on anti-$\mathcal{PT}$-symmetry. Here, we propose an algorithm for simulating the universal anti-$\mathcal{PT}$-symmetric system with quantum circuit. Utilizing the protocols, an oscillation of information flow is observed for the first time in our Nuclear Magnetic Resonance quantum simulator. We will show that information will recover from the environment completely when the anti-$\mathcal{PT}$-symmetry is broken, whereas no information can be retrieved in the symmetry-unbroken phase. Our work opens the gate for practical quantum simulation and experimental investigation of universal anti-$\mathcal{PT}$-symmetric system in quantum computer.

quant-ph↗

arXiv2018

Experimental realization of quantum algorithms for linear system inspired by adiabatic quantum computing

Quantum adiabatic algorithm is of vital importance in quantum computation field. It offers us an alternative approach to manipulate the system instead of quantum gate model. Recently, an interesting work arXiv:1805.10549 indicated that we can solve linear equation system via algorithm inspired by adiabatic quantum computing. Here we demonstrate the algorithm and realize the solution of 8-dimensional linear equations $A\textbf{x}=\textbf{b}$ in a 4-qubit nuclear magnetic resonance system. The result is by far the solution of maximum-dimensional linear equation with a limited number of qubits in experiments, which includes some ingenious simplifications. Our experiment provides the new possibility of solving so many practical problems related to linear equations systems and has the potential applications in designing the future quantum algorithms.

quant-ph↗

arXiv2018

Experimental Simulation of Bosonic Creation and Annihilation Operators in a Quantum Processor

The ability of implementing quantum operations plays fundamental role in manipulating quantum systems. Creation and annihilation operators which transform a quantum state to another by adding or subtracting a particle are crucial of constructing quantum description of many body quantum theory and quantum field theory. Here we present a quantum algorithm to perform them by the linear combination of unitary operations associated with a two-qubit ancillary system. Our method can realize creation and annihilation operators simultaneously in the subspace of the whole system. A prototypical experiment was performed with a 4-qubit Nuclear Magnetic Resonance processor, demonstrating the algorithm via full state tomography. The creation and annihilation operators are realized with a fidelity all above 96% and a probability about 50%. Moreover, our method can be employed to quantum random walk in an arbitrary initial state. With the prosperous development of quantum computing, our work provides a quantum control technology to implement non-unitary evolution in near-term quantum computer.

quant-ph↗

arXiv2018

A Quantum Algorithm for Solving Linear Differential Equations: Theory and Experiment

We present and experimentally realize a quantum algorithm for efficiently solving the following problem: given an $N\times N$ matrix $\mathcal{M}$, an $N$-dimensional vector $\textbf{\emph{b}}$, and an initial vector $\textbf{\emph{x}}(0)$, obtain a target vector $\textbf{\emph{x}}(t)$ as a function of time $t$ according to the constraint $d\textbf{\emph{x}}(t)/dt=\mathcal{M}\textbf{\emph{x}}(t)+\textbf{\emph{b}}$. We show that our algorithm exhibits an exponential speedup over its classical counterpart in certain circumstances. In addition, we demonstrate our quantum algorithm for a $4\times4$ linear differential equation using a 4-qubit nuclear magnetic resonance quantum information processor. Our algorithm provides a key technique for solving many important problems which rely on the solutions to linear differential equations.

quant-ph↗

arXiv2017

Revisiting the Master-Slave Architecture in Multi-Agent Deep Reinforcement Learning

Many tasks in artificial intelligence require the collaboration of multiple agents. We exam deep reinforcement learning for multi-agent domains. Recent research efforts often take the form of two seemingly conflicting perspectives, the decentralized perspective, where each agent is supposed to have its own controller; and the centralized perspective, where one assumes there is a larger model controlling all agents. In this regard, we revisit the idea of the master-slave architecture by incorporating both perspectives within one framework. Such a hierarchical structure naturally leverages advantages from one another. The idea of combining both perspectives is intuitive and can be well motivated from many real world systems, however, out of a variety of possible realizations, we highlights three key ingredients, i.e. composed action representation, learnable communication and independent reasoning. With network designs to facilitate these explicitly, our proposal consistently outperforms latest competing methods both in synthetic experiments and when applied to challenging StarCraft micromanagement tasks.

cs.AI↗

arXiv2017

Implementation of Multiparty quantum clock synchronization

The quantum clock synchronization (QCS) is to measure the time difference among the spatially separated clocks with the principle of quantum mechanics. The first QCS algorithm proposed by Chuang and Jozsa is merely based on two parties, which is further extended and generalized to the multiparty situation by Krco and Paul. They present a multiparty QCS protocol based upon W states that utilizes shared prior entanglement and broadcast of classical information to synchronize spatially separated clocks. Shortly afterwards, Ben-Av and Exman came up with an optimized multiparty QCS using Z state. In this work, we firstly report an implementation of Krco and Ben-AV multiparty QCS algorithm using a four-qubit Nuclear Magnetic Resonance (NMR). The experimental results show a great agreement with the theory and also prove Ben-AV multiparty QCS algorithm more accurate than Krco.

quant-ph↗

arXiv2017

Collaborative Deep Reinforcement Learning for Joint Object Search

We examine the problem of joint top-down active search of multiple objects under interaction, e.g., person riding a bicycle, cups held by the table, etc.. Such objects under interaction often can provide contextual cues to each other to facilitate more efficient search. By treating each detector as an agent, we present the first collaborative multi-agent deep reinforcement learning algorithm to learn the optimal policy for joint active object localization, which effectively exploits such beneficial contextual information. We learn inter-agent communication through cross connections with gates between the Q-networks, which is facilitated by a novel multi-agent deep Q-learning algorithm with joint exploitation sampling. We verify our proposed method on multiple object detection benchmarks. Not only does our model help to improve the performance of state-of-the-art active localization models, it also reveals interesting co-detection patterns that are intuitively interpretable.

cs.CV↗

arXiv2013

A lasing mechanism based on absorption boundary of gain materials

A new kind of mechanism of lasing is investigated experimentally. It is quite different from the traditional laser with cavity and the random laser with random scattering. In this mechanism, the intensity-dependent refractive index effect and thermal lensing effects of the pump beam induce a large gradient of the refractive index in the gain material, which forms a passive equivalent boundary that provides the feedback in the lasing system. A real lasing system, a liquid disk laser, is performed, it achieves 2-D omnidirectional radiation with a high efficiency of 28%, its radiation spectral property can be explained by resonant Raman scattering.

physics.optics↗