We present a target-driven navigation system to improve mapless visual navigation in indoor scenes. Our method takes a multi-view observation of a robot and a target image as inputs at each time step to provide a sequence of actions that move the robot to the target without relying on odometry or GPS at runtime. The system is learned by optimizing a combinational objective encompassing three key designs. First, we propose that an agent conceives the next observation before making an action decision. This is achieved by learning a variational generative module from expert demonstrations. We then propose predicting static collision in advance, as an auxiliary task to improve safety during navigation. Moreover, to alleviate the training data imbalance problem of termination action prediction, we also introduce a target checking module to differentiate from augmenting navigation policy with a termination action. The three proposed designs all contribute to the improved training data efficiency, static collision avoidance, and navigation generalization performance, resulting in a novel target-driven mapless navigation system. Through experiments on a TurtleBot, we provide evidence that our model can be integrated into a robotic system and navigate in the real world. Videos and models can be found in the supplementary material.

To enhance the cross-target and cross-scene generalization of target-driven visual navigation based on deep reinforcement learning (RL), we introduce an information-theoretic regularization term into the RL objective. The regularization maximizes the mutual information between navigation actions and visual observation transforms of an agent, thus promoting more informed navigation decisions. This way, the agent models the action-observation dynamics by learning a variational generative model. Based on the model, the agent generates (imagines) the next observation from its current observation and navigation target. This way, the agent learns to understand the causality between navigation actions and the changes in its observations, which allows the agent to predict the next action for navigation by comparing the current and the imagined next observations. Cross-target and cross-scene evaluations on the AI2-THOR framework show that our method attains at least 10% improvement of average success rate over some state-of-the-art models. We further evaluate our model in two real-world settings: navigation in unseen indoor scenes from a discrete Active Vision Dataset (AVD) and continuous real-world environments with a TurtleBot. We demonstrate that our navigation model is able to successfully achieve navigation tasks in these scenarios. ¹¹ [Online]. Available: https://github.com/wqynew/RL-based-navigation.

The 6-D pose estimation is a crucial task in vision-based measurement for robotic manipulation. It becomes a challenging task because of the variety of lighting conditions, cluttered background, occlusion, and texture-less objects. The various lighting conditions and texture-less objects lead to dramatic changes for imaging. In this article, we propose an edge-attention 6-D pose estimation network (EANet) for texture-less objects that achieve the autonomous perception of the edge, which is invariant when the lighting conditions change and objects are texture-less. To achieve this, EANet adopts a multitask learning strategy that introduces edge cues into the pixelwise dense fusion framework. We design a shared-weight edge extractor serving for edge reconstruction and pose estimation simultaneously. The purpose of edge reconstruction is to guide the network to pay more attention to edges, so as to enhance the performance of pose estimation implicitly. The edge cues can further improve the performance of pose refinement. Furthermore, we apply a skip-connection scheme to the edge extractor, merging feature maps of the deep and shallow layers. Due to the lightweight design, EANet obtains an inference time of 20 frames per second (FPS) on a desktop PC with an RTX 2070 GPU, which satisfies the requirement of most real-time applications. Experiments show that our network has a significant effect on texture-less objects and approaches the state-of-the-art performance on both the LINEMOD and the YCB-Video datasets. Finally, we present a PARTS dataset to test real-world performance and apply EANet to bin-picking vision measurement, which demonstrates that our method can provide a complete and robust solution for vision-based pose measurement.

3D pose estimation for texture-less objects remains a challenging problem. Previous works either focus on a template matching method to find the nearest template as a candidate, or construct a Hough Forest, which utilizes the offset of patches to vote for the object location and pose. By contrast, in this paper, we propose a comprehensive framework to directly regress 3D poses for the candidates, in which a CNN-based triplet network is trained to extract discriminated features from the binary images. To make the features suitable for the regression task, a pose-guided method and a regression constraint are employed to the constructed triplet network. We show that the constraint reaches the goal of creating the correlation between the features and 3D poses. Once the expected features are obtained, the object pose could be efficiently regressed by training a regression network with a simple structure. For symmetric objects, depth images are treated as an additional channel to feed the triplet network. Experiments on the LineMOD and our own datasets demonstrate our method with high regression precision and efficiency.

We present a novel approach for image-goal navigation, where an agent navigates with a goal image rather than accurate target information, which is more challenging. Our goal is to decouple the learning of navigation goal planning, collision avoidance, and navigation ending prediction, which enables more concentrated learning of each part. This is realized by four different modules. The first module maintains an obstacle map during robot navigation. The second predicts a long-term goal on the real-time map periodically, which can thus convert an image-goal navigation task to several point-goal navigation tasks. To achieve these point-goal navigation tasks, the third module plans collision-free command sets for navigating to these long-term goals. The final module stops the robot properly near the goal image. The four modules are designed or maintained separately, which helps cut down the search time during navigation and improve the generalization to previously unseen real scenes. We evaluate the method in both a simulator and in the real world with a mobile robot. The results in real complex environments show that our method attains at least a 17% increase in navigation success rate and a 23% decrease in navigation collision rate over some state-of-the-art models.

3D single object tracking is the task of localizing a target object in a search point cloud frame. In this paper, we present a multi-level structure-enhanced tracking model to improve the tracking performance in sparse 3D point clouds. Towards this end, we first encode the target and the search point clouds efficiently in two near-neighbors graphs, which allows structural information to flow between neighbors along graph edges. We then design a cross-graph attention mechanism, which associates similar nodes across the target graph and the search graph, and further dissimilar graph nodes for apart. Integrating the proposed mechanism into the above Siamese feature learning of both the target and the search frame, we strengthen the structural correlation between the target and the search frame. In that case, distinguishing the potential target from the background in the search frame would be much simpler. Finally, we design a U-shaped sparse convolutional block to aggregate the structural features of the potential target in the search frame. Integrating the proposed block into an existing target localization module from [1], we localize target centers accurately. Experiments on the KITTI benchmark demonstrate that our method outperforms some state-of-the-art models, achieving at least a $3.2\%$ improvement in terms of average tracking precision.