Yi Wu

Yunfei Li, Tao Kong, Lei Li et al. · bytedance, cmu

Can a robot autonomously learn to design and construct a bridge from varying-sized blocks without a blueprint? It is a challenging task with long horizon and sparse reward -- the robot has to figure out physically stable design schemes and feasible actions to manipulate and transport blocks. Due to diverse block sizes, the state space and action trajectories are vast to explore. In this paper, we propose a hierarchical approach for this problem. It consists of a reinforcement-learning designer to propose high-level building instructions and a motion-planning-based action generator to manipulate blocks at the low level. For high-level learning, we develop a novel technique, prioritized memory resetting (PMR) to improve exploration. PMR adaptively resets the state to those most critical configurations from a replay buffer so that the robot can resume training on partial architectures instead of from scratch. Furthermore, we augment PMR with auxiliary training objectives and fine-tune the designer with the locomotion generator. Our experiments in simulation and on a real deployed robotic system demonstrate that it is able to effectively construct bridges with blocks of varying sizes at a high success rate. Demos can be found at https://sites.google.com/view/bridge-pmr.

Jiayi Chen, Ruifeng Gao, Jue Wang et al.

The construction of channel gain map (CGM) is essential for realizing environment-aware wireless communications expected in 6G, for which a fundamental problem is how to predict the channel gains at unknown locations effectively by a finite number of measurements. As using a single prediction model is not effective in complex propagation environments, we propose a subregional learning-based CGM construction scheme, with which the entire map is divided into subregions via data-driven clustering, then individual models are constructed and trained for every subregion. In this way, specific propagation feature in each subregion can be better extracted with finite training data. Moreover, we propose to further improve prediction accuracy by uneven subregion sampling, as well as training data reuse around the subregion boundaries. Simulation results validate the effectiveness of the proposed scheme in CGM construction.

Zhe Zhang, Shiyao Ma, Zhaohui Yang et al.

Air access networks have been recognized as a significant driver of various Internet of Things (IoT) services and applications. In particular, the aerial computing network infrastructure centered on the Internet of Drones has set off a new revolution in automatic image recognition. This emerging technology relies on sharing ground truth labeled data between Unmanned Aerial Vehicle (UAV) swarms to train a high-quality automatic image recognition model. However, such an approach will bring data privacy and data availability challenges. To address these issues, we first present a Semi-supervised Federated Learning (SSFL) framework for privacy-preserving UAV image recognition. Specifically, we propose model parameters mixing strategy to improve the naive combination of FL and semi-supervised learning methods under two realistic scenarios (labels-at-client and labels-at-server), which is referred to as Federated Mixing (FedMix). Furthermore, there are significant differences in the number, features, and distribution of local data collected by UAVs using different camera modules in different environments, i.e., statistical heterogeneity. To alleviate the statistical heterogeneity problem, we propose an aggregation rule based on the frequency of the client's participation in training, namely the FedFreq aggregation rule, which can adjust the weight of the corresponding local model according to its frequency. Numerical results demonstrate that the performance of our proposed method is significantly better than those of the current baseline and is robust to different non-IID levels of client data.

Xiaoteng Zhou, Changli Yu, Xin Yuan et al.

With the continuous development of underwater vision technology, more and more remote sensing images could be obtained. In the underwater scene, sonar sensors are currently the most effective remote perception devices, and the sonar images captured by them could provide rich environment information. In order to analyze a certain scene, we often need to merge the sonar images from different periods, various sonar frequencies and distinctive viewpoints. However, the above scenes will bring nonlinear intensity differences to the sonar images, which will make traditional matching methods almost ineffective. This paper proposes a non-linear intensity sonar image matching method that combines local feature points and deep convolution features. This method has two key advantages: (i) we generate data samples related to local feature points based on the self-learning idea; (ii) we use the convolutional neural network (CNN) and Siamese network architecture to measure the similarity of the local position in the sonar image pair. Our method encapsulates the feature extraction and feature matching stage in a model, and directly learns the mapping function from image patch pairs to matching labels, and achieves matching tasks in a near-end-to-end manner. Feature matching experiments are carried out on the sonar images acquired by autonomous underwater vehicle (AUV) in the real underwater environment. Experiment results show that our method has better matching effects and strong robustness.