Multi-modal fusion of sensors is a commonly used approach to enhance the performance of odometry estimation, which is also a fundamental module for mobile robots. However, the question of \textit{how to perform fusion among different modalities in a supervised sensor fusion odometry estimation task?} is still one of challenging issues remains. Some simple operations, such as element-wise summation and concatenation, are not capable of assigning adaptive attentional weights to incorporate different modalities efficiently, which make it difficult to achieve competitive odometry results. Recently, the Transformer architecture has shown potential for multi-modal fusion tasks, particularly in the domains of vision with language. In this work, we propose an end-to-end supervised Transformer-based LiDAR-Inertial fusion framework (namely TransFusionOdom) for odometry estimation. The multi-attention fusion module demonstrates different fusion approaches for homogeneous and heterogeneous modalities to address the overfitting problem that can arise from blindly increasing the complexity of the model. Additionally, to interpret the learning process of the Transformer-based multi-modal interactions, a general visualization approach is introduced to illustrate the interactions between modalities. Moreover, exhaustive ablation studies evaluate different multi-modal fusion strategies to verify the performance of the proposed fusion strategy. A synthetic multi-modal dataset is made public to validate the generalization ability of the proposed fusion strategy, which also works for other combinations of different modalities. The quantitative and qualitative odometry evaluations on the KITTI dataset verify the proposed TransFusionOdom could achieve superior performance compared with other related works.
相關內容
Structural magnetic resonance imaging (sMRI) has shown great clinical value and has been widely used in deep learning (DL) based computer-aided brain disease diagnosis. Previous approaches focused on local shapes and textures in sMRI that may be significant only within a particular domain. The learned representations are likely to contain spurious information and have a poor generalization ability in other diseases and datasets. To facilitate capturing meaningful and robust features, it is necessary to first comprehensively understand the intrinsic pattern of the brain that is not restricted within a single data/task domain. Considering that the brain is a complex connectome of interlinked neurons, the connectional properties in the brain have strong biological significance, which is shared across multiple domains and covers most pathological information. In this work, we propose a connectional style contextual representation learning model (CS-CRL) to capture the intrinsic pattern of the brain, used for multiple brain disease diagnosis. Specifically, it has a vision transformer (ViT) encoder and leverages mask reconstruction as the proxy task and Gram matrices to guide the representation of connectional information. It facilitates the capture of global context and the aggregation of features with biological plausibility. The results indicate that CS-CRL achieves superior accuracy in multiple brain disease diagnosis tasks across six datasets and three diseases and outperforms state-of-the-art models. Furthermore, we demonstrate that CS-CRL captures more brain-network-like properties, better aggregates features, is easier to optimize and is more robust to noise, which explains its superiority in theory. Our source code will be released soon.
When introducing physics-constrained deep learning solutions to the volumetric super-resolution of scientific data, the training is challenging to converge and always time-consuming. We propose a new hierarchical sampling method based on octree to solve these difficulties. In our approach, scientific data is preprocessed before training, and a hierarchical octree-based data structure is built to guide sampling on the latent context grid. Each leaf node in the octree corresponds to an indivisible subblock of the volumetric data. The dimensions of the subblocks are different, making the number of sample points in each randomly cropped training data block to be adaptive. We reconstruct the octree at intervals according to loss distribution to perform the multi-stage training. With the Rayleigh-B\'enard convection problem, we deploy our method to state-of-the-art models. We constructed adequate experiments to evaluate the training performance and model accuracy of our method. Experiments indicate that our sampling optimization improves the convergence performance of physics-constrained deep learning super-resolution solutions. Furthermore, the sample points and training time are significantly reduced with no drop in model accuracy. We also test our method in training tasks of other deep neural networks, and the results show our sampling optimization has extensive effectiveness and applicability. The code is publicly available at //github.com/xinjiewang/octree-based_sampling.
Nonhomogeneous partial differential equations (PDEs) are an applicable model in soft sensor modeling for describing spatiotemporal industrial systems with unmeasurable source terms, which cannot be well solved by existing physics-informed neural networks (PINNs). To this end, a coupled PINN (CPINN) with a recurrent prediction (RP) learning strategy (CPINN-RP) is proposed for soft sensor modeling in spatiotemporal industrial processes, such as vibration displacement. First, CPINN containing NetU and NetG is proposed. NetU is used to approximate the solutions to PDEs under study and NetG is used to regularize the training of NetU. The two networks are integrated into a data-physics-hybrid loss function. Then, we theoretically prove that the proposed CPINN has a satisfying approximation capacity to the PDEs solutions. Besides the theoretical aspects, we propose a hierarchical training strategy to optimize and couple the two networks to achieve the parameters of CPINN. Secondly, NetU-RP is achieved by NetU compensated by RP, the recurrently delayed output of CPINN, to further improve the soft sensor performance. Finally, simulations and experiment verify the effectiveness and practical applications of CPINN-RP.
Accurately detecting lane lines in 3D space is crucial for autonomous driving. Existing methods usually first transform image-view features into bird-eye-view (BEV) by aid of inverse perspective mapping (IPM), and then detect lane lines based on the BEV features. However, IPM ignores the changes in road height, leading to inaccurate view transformations. Additionally, the two separate stages of the process can cause cumulative errors and increased complexity. To address these limitations, we propose an efficient transformer for 3D lane detection. Different from the vanilla transformer, our model contains a decomposed cross-attention mechanism to simultaneously learn lane and BEV representations. The mechanism decomposes the cross-attention between image-view and BEV features into the one between image-view and lane features, and the one between lane and BEV features, both of which are supervised with ground-truth lane lines. Our method obtains 2D and 3D lane predictions by applying the lane features to the image-view and BEV features, respectively. This allows for a more accurate view transformation than IPM-based methods, as the view transformation is learned from data with a supervised cross-attention. Additionally, the cross-attention between lane and BEV features enables them to adjust to each other, resulting in more accurate lane detection than the two separate stages. Finally, the decomposed cross-attention is more efficient than the original one. Experimental results on OpenLane and ONCE-3DLanes demonstrate the state-of-the-art performance of our method.
Humans naturally exploit haptic feedback during contact-rich tasks like loading a dishwasher or stocking a bookshelf. Current robotic systems focus on avoiding unexpected contact, often relying on strategically placed environment sensors. Recently, contact-exploiting manipulation policies have been trained in simulation and deployed on real robots. However, they require some form of real-world adaptation to bridge the sim-to-real gap, which might not be feasible in all scenarios. In this paper we train a contact-exploiting manipulation policy in simulation for the contact-rich household task of loading plates into a slotted holder, which transfers without any fine-tuning to the real robot. We investigate various factors necessary for this zero-shot transfer, like time delay modeling, memory representation, and domain randomization. Our policy transfers with minimal sim-to-real gap and significantly outperforms heuristic and learnt baselines. It also generalizes to plates of different sizes and weights. Demonstration videos and code are available at //sites.google.com/view/compliant-object-insertion.
As the reliability of the robot's perception correlates with the number of integrated sensing modalities to tackle uncertainty, a practical solution to manage these sensors from different computers, operate them simultaneously, and maintain their real-time performance on the existing robotic system with minimal effort is needed. In this work, we present an end-to-end software-hardware framework, namely ExtPerFC, that supports both conventional hardware and software components and integrates machine learning object detectors without requiring an additional dedicated graphic processor unit (GPU). We first design our framework to achieve real-time performance on the existing robotic system, guarantee configuration optimization, and concentrate on code reusability. We then mathematically model and utilize our transfer learning strategies for 2D object detection and fuse them into depth images for 3D depth estimation. Lastly, we systematically test the proposed framework on the Baxter robot with two 7-DOF arms, a four-wheel mobility base, and an Intel RealSense D435i RGB-D camera. The results show that the robot achieves real-time performance while executing other tasks (e.g., map building, localization, navigation, object detection, arm moving, and grasping) simultaneously with available hardware like Intel onboard CPUS/GPUs on distributed computers. Also, to comprehensively control, program, and monitor the robot system, we design and introduce an end-user application. The source code is available at //github.com/tuantdang/perception_framework.
We propose a sensory substitution device that communicates one-degree-of-freedom proprioceptive feedback via deep pressure stimulation on the arm. The design is motivated by the need for a feedback modality detectable by individuals with a genetic condition known as PIEZO2 loss of function, which is characterized by absence of both proprioception and sense of light touch. We created a wearable and programmable prototype that applies up to 15 N of deep pressure stimulation to the forearm and includes an embedded force sensor. We conducted a study to evaluate the ability of participants without sensory impairment to control the position of a virtual arm to match a target angle communicated by deep pressure stimulation. A participant-specific calibration resulted in an average minimum detectable force of 0.41 N and maximum comfortable force of 6.42 N. We found that, after training, participants were able to significantly reduce angle error using the deep pressure haptic feedback compared to without it. Angle error increased only slightly with force, indicating that this sensory substitution method is a promising approach for individuals with PIEZO2 loss of function and other forms of sensory loss.
This work investigates the use of a Deep Neural Network (DNN) to perform an estimation of the Weapon Engagement Zone (WEZ) maximum launch range. The WEZ allows the pilot to identify an airspace in which the available missile has a more significant probability of successfully engaging a particular target, i.e., a hypothetical area surrounding an aircraft in which an adversary is vulnerable to a shot. We propose an approach to determine the WEZ of a given missile using 50,000 simulated launches in variate conditions. These simulations are used to train a DNN that can predict the WEZ when the aircraft finds itself on different firing conditions, with a coefficient of determination of 0.99. It provides another procedure concerning preceding research since it employs a non-discretized model, i.e., it considers all directions of the WEZ at once, which has not been done previously. Additionally, the proposed method uses an experimental design that allows for fewer simulation runs, providing faster model training.
This work addresses a novel and challenging problem of estimating the full 3D hand shape and pose from a single RGB image. Most current methods in 3D hand analysis from monocular RGB images only focus on estimating the 3D locations of hand keypoints, which cannot fully express the 3D shape of hand. In contrast, we propose a Graph Convolutional Neural Network (Graph CNN) based method to reconstruct a full 3D mesh of hand surface that contains richer information of both 3D hand shape and pose. To train networks with full supervision, we create a large-scale synthetic dataset containing both ground truth 3D meshes and 3D poses. When fine-tuning the networks on real-world datasets without 3D ground truth, we propose a weakly-supervised approach by leveraging the depth map as a weak supervision in training. Through extensive evaluations on our proposed new datasets and two public datasets, we show that our proposed method can produce accurate and reasonable 3D hand mesh, and can achieve superior 3D hand pose estimation accuracy when compared with state-of-the-art methods.
Multimodal sentiment analysis is a very actively growing field of research. A promising area of opportunity in this field is to improve the multimodal fusion mechanism. We present a novel feature fusion strategy that proceeds in a hierarchical fashion, first fusing the modalities two in two and only then fusing all three modalities. On multimodal sentiment analysis of individual utterances, our strategy outperforms conventional concatenation of features by 1%, which amounts to 5% reduction in error rate. On utterance-level multimodal sentiment analysis of multi-utterance video clips, for which current state-of-the-art techniques incorporate contextual information from other utterances of the same clip, our hierarchical fusion gives up to 2.4% (almost 10% error rate reduction) over currently used concatenation. The implementation of our method is publicly available in the form of open-source code.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191