Autonomous driving technology is becoming a significant trend in the development of public transportation. A critical task in autonomous driving perception is 3D object detection, which provides essential data support for downstream applications. Most mainstream 3D object detection methods rely on the Cartesian coordinate system, where they construct object queries to interact with image features and position embedding. However, these methods have the following problems: 1) Sensor-captured detail information diminishes with increasing distance, while pixels represent the same space in Cartesian coordinates, preventing the model from fully leveraging details in closer regions. 2) Multi-view images suffer from spatial misalignment due to overlapping fields of view. 3) The performance of existing single-branch depth prediction networks lacks the necessary accuracy. These issues hinder the feature interaction and affect detection performance. We propose an innovative framework PRTF. Based on Polar space, we design the Two-Stage Transformation Encoder: in the first stage, Dual-DepthNet is used to improve the accuracy of depth prediction. In the second stage, Polar points are generated to address spatial misalignment, enabling effective encoding of details at close distance. In the Temporal Decoder, object queries are leveraged to integrate temporal information, effectively compensating for ambiguous information. By enhancing spatial information at both near and far distances in Polar space, the overall performance of multi-view 3D object detection is significantly improved. PRTF achieves state-of-the-art performance on nuScenes Test with 56.1% mAP and 63.9% NDS, exceeding multi-modal frameworks that combine image and radar data.

Sparse Query-Based paradigms for multi-view 3D object detection achieved remarkable success, but the redundant predictions still affects the detection accuracy and object localization because of the following shortcomings: 1) 3D position embedding exhibits weak spatial perception ability and cannot capture subtle differences between similar objects. 2) Randomly initialized object queries lack prior knowledge of reference points and object information, which hinders their accuracy in matching with corresponding objects. 3) The temporal fusion process fails to effectively capture object details, making it challenging to localize object accurately. To address these issues, we propose an innovative framework T-Match, which leverages prior knowledge from a 2D detector to initialize object queries. Based on Polar space, both 2D and 3D information are integrated to comprehensively capture object details, object queries iteratively updates through Cross-Domain Spatio-Temporal Attention, which incorporates cross-domain object information, and Polar-Aware Cross Attention, which aggregates image features fused with Polar position embedding, refining matching results while reducing redundant predictions. T-Match achieves state-of-the-art performance on nuScenes Test with 57.5% mAP and 64.9% NDS, exceeding multi-modal frameworks that combine image and radar data.

From the experience of offloading the network function of elastic IP (EIP) gateway to programming protocol independent packet processors (P4) switches at the edge cloud, we analyze the challenges of limited on-chip resources and share our insights to ensure the gateway’s quality of service (QoS). In this paper, we propose the design of EIP traffic management based on a P4 virtual buffer (VBuf), which makes use of surplus throughput. The novel design is implemented entirely on the network data plane in P4 and combines the advantages of general traffic shaping (GTS) and committed access rate (CAR). Then, evaluation experiments are carried out to verify QoS improvements. Their results show that the proposed method can reduce local traffic jitter by 4X and local packet loss by 10X in comparison with CAR. Besides, the proposed method leads to smoother congestion window change for TCP traffic.

IT services are now popularly hosted in cloud systems. In order to enhance the availability of cloud services, an emerging approach for detecting failures of cloud components is to monitor Key Performance Indicators (KPIs) of the components and apply Neural Network based AI technologies to detect KPI anomalies. Multivariate Time Series Anomaly Detection (TSAD) models have been designed for this purpose. However, when applying such models directly to real-world cloud systems the anomaly detection performance is not as good. This is because the number of KPIs in real cloud systems is typically much more than the number of KPIs in the datasets used for model evaluation, and the larger number of KPIs bring about a performance loss of the models’ anomaly detection. Therefore, selecting KPIs properly is essential for applying multivariant KPI data for any practical anomaly detection. This paper studies this performance loss issue when TSAD models are applied onto real-world cloud systems, and proposes fKPISelect, a mechanism of automated KPI selection based on fault injection. We implemented fKPISelect, deployed it to a real cloud system, and created a real-world KPI dataset. We conducted extensive experiments, and the experimental results show the effectiveness and practicality of fKPISelect: it improves the F1 score of anomaly detection from 0.68 to 0.91 for real-world KPI data.