In this paper, a hierarchical control scheme, including a decision-making layer and a control layer, is proposed for vehicle platoons executing vehicle-following, cut-in, and cut-out maneuvers. Based on a finite-state machine, the decision-making layer is designed to ensure collision avoidance for vehicle platoons. In the control layer, a distributed model predictive control strategy is employed in the outer-loop to generate a reference sequence, where a linear parameter-varying kinematic model is established to account for the coupling of the vehicle platoon. Incremental constraints are designed to ensure bumpless transfer control during phase switching. By selecting the sum of local cost functions as a Lyapunov candidate, the asymptotic consensus of the vehicle platoon is proven. Furthermore, considering the coupled longitudinal and lateral dynamics of vehicles, in the inner-loop, a nonlinear model predictive control strategy is proposed to track the reference sequence. The effectiveness of the hierarchical control scheme is validated through co-simulation using MATLAB and TruckSim.

In this paper, a hierarchical control strategy of vehicle platoons is presented, in which the longitudinal and lateral coupling property of vehicles is taken into account. A three-degree-of-freedom dynamic model of vehicles is approximated to a “global” linear model by the Koopman operator theory. A synchronous distributed predictive control scheme of vehicle platoons is proposed as an upper-level controller, where both the linear vehicle model and a linear parametric-varying lane-keeping model are adopted to predict the dynamic of vehicles, and keep vehicles in the designated lane. Thus, it can avoid the solution of nonlinear optimization problems and reduce the computational burden accordingly. A lower-level controller is designed, where the desired longitudinal control force determined by the upper-level controller is transformed into the desired throttle angle and brake pressure through an inverse longitudinal dynamics model of vehicles. The joint simulation results by PreScan, CarSim and MATLAB/Simulink show that when the leader vehicle accelerates or decelerates, the following vehicles in the platoon can keep the same velocity as the leader vehicle, and maintain the desired safety distance between the front and rear vehicles. In addition, joint simulation in the curved road scenario show that the performance of lane keeping can be guaranteed for vehicle platoons with the proposed control strategy.

Fine-tuning on cheap commodity GPU servers makes large-scale deep learning models benefit more people. However, the low inter-GPU communication bandwidth and pressing communication contention on the commodity GPU server obstruct training efficiency.

Embedding models are effective for learning high-dimensional sparse data. Traditionally, they are deployed in DRAM parameter servers (PS) for online inference access. However, the ever-increasing model capacity makes this practice suffer from both high storage costs and long recovery time. Rapidly developing Persistent Memory (PM) offers new opportunities to PSs owing to its large capacity at low costs, as well as its persistence, while the application of PM also faces two challenges including high read latency and heavy CPU burden. To provide a low-cost but still high-performance parameter service for online inferences, we introduce PetPS, the first production-deployed PM parameter server. (1) To escape with high PM latency, PetPS introduces a PM hash index tailored for embedding model workloads, to minimize PM access. (2) To alleviate the CPU burden, PetPS offloads parameter gathering to NICs, to avoid CPU stalls when accessing parameters on PM and thus improve CPU efficiency. Our evaluation shows that PetPS can boost throughput by 1.3 -- 1.7X compared to PSs that use state-of-the-art PM hash indexes, or get 2.9 -- 5.5X latency reduction with the same throughput. Since 2020, PetPS has been deployed in Kuaishou, one world-leading short video company, and successfully reduced TCO by 30% without performance degradation.

In this paper, we proposed new framework for human action representation, which leverages the strengths of convolutional neural networks (CNNs) and the linear dynamical system (LDS) to represent both spatial and temporal structures of actions in videos. We make two principal contributions: first, we incorporate image-trained CNNs to detect action clip concepts, which takes advantage of different levels of information by combining the two layers in CNNs trained from images; Second, we further propose adopting a linear dynamical system (LDS) to model the relationships between these clip concepts, which captures temporal structures of actions. We have applied the proposed method on two challenging realistic benchmark datasets, and our method achieves high performance up to 86.16% on the YouTube and 82.76% UCF50 datasets, which largely outperforms most of the state-of-the-art algorithms with more sophisticated techniques.

Human action recognition from realistic scenarios is extremely challenging due to large intra-class variation and complex background clutters. In this paper, by leveraging the strength of deep learning and vector of locally aggregated descriptors (VLAD), we propose a new methods for human action recognition from realistic datsets. We adopt stack convolu-tional independent subspace analysis (ISA) networks to learn 3D cuboid representation directly from spatio-temporal video data; we propose an improved VLAD by incorporating the spatio-temporal geometrical information to encode the deep learned local features. On two challenging realistic datasets: the YouTube action and HMDB51 datasets, the proposed method achieves state-of-the-art performance with an efficient linear SVM classifier, which is competitive with and even better than existing sophisticated algorithms.