Jinzhi Wang

This paper investigates the distributed consensus control problem for linear multiagent systems with unknown system parameters. First, a time-varying estimation system is designed to dynamically estimate the unknown parameters. At each sampling instant, an iterative estimation algorithm is developed to update the parameter estimates using state errors and control inputs. Based on the estimation system, a distributed model predictive control (MPC) algorithm is proposed to achieve consensus. The Lyapunov stability theory is employed to rigorously prove consensus convergence for both the estimation system and the original multiagent system. Finally, the complete distributed MPC algorithm is presented and its effectiveness is demonstrated through a numerical simulation.

Constrained robots play an important role in industrial processes and delivery tasks. To improve the numerical accuracy of robot state estimation, many filtering methods including the well-known unscented Kalman filter (UKF) have been widely studied. However, most conventional propagation schemes in UKF are based on a direct discretization of the continuous-time equations, which suffer from the problem of ignoring physical structures and properties of a robot (such as physical constraints, energy conservation, and manifold structure preservation) due to the numerical dissipation of the time discretization scheme. In this letter, we introduce a structure-preserving unscented Kalman filter (SP-UKF) for mobile robots. By using differential geometry, the resulting time propagation step of the proposed filter shows the benefit of preserving the no-slip constraint of a mobile robot and at the same time respecting key structures and physical laws of the system. Numerical results validate the efficiency and the structure-preserving properties of the proposed approach.

This paper investigates the design of distributed observers for a class of nonlinear systems. The designed distributed observers reside in a network of sensor nodes. The communication links in the network enable each sensor node to estimate all states of the nonlinear system, even though each one can only collect a small portion of the system's outputs. However, all inputs of the system are still needed at each node to carry out a correct state estimation. For this problem, the paper further proposes a distributed nonlinear observer without using global inputs, in which each sensor node only needs partial input information of the nonlinear system.

The goal of conditional image-to-video (cI2V) generation is to create a believable new video by beginning with the condition, i.e., one image and text. The previous cI2V generation methods conventionally perform in RGB pixel space, with limitations in modeling motion consistency and visual continuity. Additionally, the efficiency of generating videos in pixel space is quite low. In this paper, we propose a novel approach to address these challenges by disentangling the target RGB pixels into two distinct components: spatial content and temporal motions. Specifically, we predict temporal motions which include motion vector and residual based on a 3D-UNet diffusion model. By explicitly modeling temporal motions and warping them to the starting image, we improve the temporal consistency of generated videos. This results in a reduction of spatial redundancy, emphasizing temporal details. Our proposed method achieves performance improvements by disentangling content and motion, all without introducing new structural complexities to the model. Extensive experiments on various datasets confirm our approach's superior performance over the majority of state-of-the-art methods in both effectiveness and efficiency.