We present BehaviorKit, an open-source plug-and-play module that enables the addition of real-time behavior analysis using deep learning to any robot with internet access. The library bundles together gaze tracking, real-time transcription, textual sentiment analysis, facial emotion (valence and arousal) estimation, and face and pose landmark localization. The library is designed to be run on a GPU-powered computer, either on the robot or externally, to process the incoming visual and auditory inputs that the robot receives. The library connects via WebSocket to the robot, which receives the processed outputs from all of the models. The WebSocket client can easily connect to a ROS-powered robot, or a custom client can be written to adapt to any robot; alternatively, the library can also be run between two laptop computers. A small dataset is provided to test the framework. By packaging together and optimizing many commonly used models, we hope to enable easier access to high-performing behavior models for the HRI community. The code is available at https://github.com/IUB-RHouse/BehaviorKit.

Vehicle-to-Vehicle (V2V) cooperative perception has become increasingly popular in the field of autonomous driving, effectively overcoming the inherent limitations of single-vehicle perception systems, such as limited range and susceptibility to occlusions. In a V2V system, vehicles in close proximity can share perception data. To fuse this data, which is collected from different viewpoints by each vehicle, accurate pose information (including position and heading direction) is essential to transform the received data to the receiving vehicle's viewpoint. However, pose errors, often caused by measurement noise or sensor failures, can lead to severe misalignment during data fusion, resulting in incorrect object detections and potentially hazardous decisions in autonomous driving systems. To address this challenge, we present BB-Align, a lightweight pose recovery framework that utilizes Lidar Bird's-eye View (BV) images and object bounding Boxes for relative pose estimation. Designed as a plug-and-play solution, the proposed method requires no additional model training, enabling effortless integration into existing V2V systems. Our approach uses Lidar-derived BV images with a Log-Gabor filter-based feature map for effective image matching despite image sparsity. To reduce errors from self-motion distortion, we also integrate object bounding boxes for finer alignment. The proposed method is rigorously evaluated on the V2V 4Real dataset-currently the only real-world V2V dataset. Our approach demonstrates high pose estimation accuracy, outperforming an existing graph-matching method. It achieves translation and rotation errors of less than 1 m and 1°, respectively, in 80 % of cases within a 70 m range between vehicles. Furthermore, by integrating the proposed framework into cooperative object detection models under serious pose error, the result shows up to a 2x increase in Average Precision (AP) compared to those without pose recovery, with more pronounced improvements in the short range.