Currently, multi-modal 3D object detection methods have become a key area of research in the field of autonomous driving. Fusion is an essential factor affecting performance in multi-modal object detection. However, previous methods still suffer from the inability to effectively fuse features from LiDAR and RGB images, resulting in a low utilization rate of complementary information between depth and semantic texture features. At the same time, existing methods may not adequately capture the structural information in Region of Interest (RoI) features when extracting them. Structural information plays a crucial role in RoI features. It encompasses the position, size, and orientation of objects, as well as the relative positions and spatial relationships between objects. Its absence can result in false or missed detections. To solve the above problems, we propose a multi-modal sensor fusion network, Bi-Att3DDet, which mainly consists of a Self-Attentive RoI Feature Extraction module (SARoIFE) and a Feature Bidirectional Interactive Fusion module (FBIF). Specifically, SARoIFE captures the relationship between different positions in RoI features to obtain high-quality RoI features through the self-attention mechanism. SARoIFE prepares for the fusion stage. FBIF performs bidirectional interaction between LiDAR and pseudo RoI features to make full use of the complementary information. We perform comprehensive experiments on the KITTI dataset, and our method notably demonstrates a 1.55% improvement in the hard difficulty level and a 0.19% improvement in the mean Average Precision (mAP) metric on the test dataset.

Autonomous driving pays considerable attention to pedestrian trajectory prediction as a crucial task. Constructing effective pedestrian trajectory prediction models depends heavily on utilizing the motion characteristics of pedestrians, along with their interactions among themselves and between themselves and their environment. However, traditional trajectory prediction models often fall short of capturing complex real-world scenarios. To address these challenges, this paper proposes an enhanced pedestrian trajectory prediction model, M²Tames, which incorporates comprehensive motion, interaction, and semantic context factors. M²Tames provides an interaction module (IM), which consists of an improved multi-head mask temporal attention mechanism (M²Tea) and an Interaction Inference Module (I²). M²Tea thoroughly characterizes the historical trajectories and potential interactions, while I² determines the precise interaction types. Then, IM adaptively aggregates useful neighbor features to generate a more accurate interactive feature map and feeds it into the final layer of the U-Net encoder to fuse with the encoder’s output. Furthermore, by adopting the U-Net architecture, M²Tames can learn and interpret scene semantic information, enhancing its understanding of the spatial relationships between pedestrians and their surroundings. These innovations improve the accuracy and adaptability of the model for predicting pedestrian trajectories. Finally, M²Tames is evaluated on the ETH/UCY and SDD datasets for short- and long-term settings, respectively. The results demonstrate that M²Tames outperforms the state-of-the-art model MSRL by 2.49% (ADE) and 8.77% (FDE) in the short-term setting and surpasses the optimum Y-Net by 6.89% (ADE) and 1.12% (FDE) in the long-term prediction. Excellent performance is also shown on the ETH/UCY datasets.