Identifying an appropriate task space can simplify solving robotic manipulation problems. One solution is deploying control algorithms in a learned low-dimensional action space. Linear and nonlinear action mapping methods have trade-offs between simplicity and the ability to express motor commands outside of a single low-dimensional subspace. We propose that learning local linear action representations can achieve both of these benefits. Our state-conditioned linear maps ensure that for any given state, the high-dimensional robotic actuation is linear in the low-dimensional actions. As the robot state evolves, so do the action mappings, so that necessary motions can be performed during a task. These local linear representations guarantee desirable theoretical properties by design. We validate these findings empirically through two user studies. Results suggest state-conditioned linear maps outperform conditional autoencoder and PCA baselines on a pick-and-place task and perform comparably to mode switching in a more complex pouring task.

A default assumption in reinforcement learning (RL) and optimal control is that observations arrive at discrete time points on a fixed clock cycle. Yet, many applications involve continuous-time systems where the time discretization, in principle, can be managed. The impact of time discretization on RL methods has not been fully characterized in existing theory, but a more detailed analysis of its effect could reveal opportunities for improving data-efficiency. We address this gap by analyzing Monte-Carlo policy evaluation for LQR systems and uncover a fundamental trade-off between approximation and statistical error in value estimation. Importantly, these two errors behave differently to time discretization, leading to an optimal choice of temporal resolution for a given data budget. These findings show that managing the temporal resolution can provably improve policy evaluation efficiency in LQR systems with finite data. Empirically, we demonstrate the trade-off in numerical simulations of LQR instances and standard RL benchmarks for non-linear continuous control.

Cross-Entropy Method (CEM) is commonly used for planning in model-based reinforcement learning (MBRL) where a centralized approach is typically utilized to update the sampling distribution based on only the top-$k$ operation's results on samples. In this paper, we show that such a centralized approach makes CEM vulnerable to local optima, thus impairing its sample efficiency. To tackle this issue, we propose Decentralized CEM (DecentCEM), a simple but effective improvement over classical CEM, by using an ensemble of CEM instances running independently from one another, and each performing a local improvement of its own sampling distribution. We provide both theoretical and empirical analysis to demonstrate the effectiveness of this simple decentralized approach. We empirically show that, compared to the classical centralized approach using either a single or even a mixture of Gaussian distributions, our DecentCEM finds the global optimum much more consistently thus improves the sample efficiency. Furthermore, we plug in our DecentCEM in the planning problem of MBRL, and evaluate our approach in several continuous control environments, with comparison to the state-of-art CEM based MBRL approaches (PETS and POPLIN). Results show sample efficiency improvement by simply replacing the classical CEM module with our DecentCEM module, while only sacrificing a reasonable amount of computational cost. Lastly, we conduct ablation studies for more in-depth analysis. Code is available at https://github.com/vincentzhang/decentCEM.

Deep learning approaches often require huge datasets to achieve good generalization. This complicates its use in tasks like image-based medical diagnosis, where the small training datasets are usually insufficient to learn appropriate data representations. For such sensitive tasks it is also important to provide the confidence in the predictions. Here, we propose a way to learn and use probabilistic labels to train accurate and calibrated deep networks from relatively small datasets. We observe gains of up to 22% in the accuracy of models trained with these labels, as compared with traditional approaches, in three classification tasks: diagnosis of hip dysplasia, fatty liver, and glaucoma. The outputs of models trained with probabilistic labels are calibrated, allowing the interpretation of its predictions as proper probabilities. We anticipate this approach will apply to other tasks where few training instances are available and expert knowledge can be encoded as probabilities.

We study the problem of learning manipulation skills from human demonstration video by inferring the association relationships between geometric features. Motivation for this work stems from the observation that humans perform eye-hand coordination tasks by using geometric primitives to define a task while a geometric control error drives the task through execution. We propose a graph based kernel regression method to directly infer the underlying association constraints from human demonstration video using Incremental Maximum Entropy Inverse Reinforcement Learning (InMaxEnt IRL). The learned skill inference provides human readable task definition and outputs control errors that can be directly plugged into traditional controllers. Our method removes the need for tedious feature selection and robust feature trackers required in traditional approaches (e.g. feature-based visual ser-voing). Experiments show our method infers correct geometric associations even with only one human demonstration video and can generalize well under variance.

Deep Convolutional Neural Networks have been adopted for salient object detection and achieved the state-of-the-art performance. Most of the previous works however focus on region accuracy but not on the boundary quality. In this paper, we propose a predict-refine architecture, BASNet, and a new hybrid loss for Boundary-Aware Salient object detection. Specifically, the architecture is composed of a densely supervised Encoder-Decoder network and a residual refinement module, which are respectively in charge of saliency prediction and saliency map refinement. The hybrid loss guides the network to learn the transformation between the input image and the ground truth in a three-level hierarchy -- pixel-, patch- and map- level -- by fusing Binary Cross Entropy (BCE), Structural SIMilarity (SSIM) and Intersection-over-Union (IoU) losses. Equipped with the hybrid loss, the proposed predict-refine architecture is able to effectively segment the salient object regions and accurately predict the fine structures with clear boundaries. Experimental results on six public datasets show that our method outperforms the state-of-the-art methods both in terms of regional and boundary evaluation measures. Our method runs at over 25 fps on a single GPU. The code is available at: https://github.com/NathanUA/BASNet.

This work describes the development of a robotic system that acquires knowledge incrementally through human interaction where new objects and motions are taught on the fly. The robotic system developed was one of the five finalists in the KUKA Innovation Award competition and demonstrated during the Hanover Messe 2018 in Germany. The main contributions of the system are i) a novel incremental object learning module - a deep learning based localization and recognition system - that allows a human to teach new objects to the robot, ii) an intuitive user interface for specifying 3D motion task associated with the new object, and iii) a hybrid force-vision control module for performing compliant motion on an unstructured surface. This paper describes the implementation and integration of the main modules of the system and summarizes the lessons learned from the competition.

We present a robot eye-hand coordination learning method that can directly learn visual task specification by watching human demonstrations. Task specification is represented as a task function, which is learned using inverse reinforcement learning(IRL [1]) by inferring a reward model from state transitions. The learned reward model is then used as continuous feedbacks in an uncalibrated visual servoing(UVS [2]) controller designed for the execution phase. Our proposed method can directly learn from raw videos, which removes the need for hand-engineered task specification. Benefiting from the use of a traditional UVS controller, the training on real robot only happens at initial Jacobian estimation which takes an average of 4-7 seconds for a new task. Besides, the learned policy is independent from a particular robot, thus has the potential of fast adapting to other robot platforms. Various experiments were designed to show that, for a task with certain DOFs, our method can adapt to task/environment changes in target positions, backgrounds, illuminations, and occlusions.

Although semi-dense Simultaneous Localization and Mapping (SLAM) has been becoming more popular over the last few years, there is a lack of efficient methods for representing and processing their large scale point clouds. In this paper, we propose using 3D line segments to simplify the point clouds generated by semi-dense SLAM. Specifically, we present a novel incremental approach for 3D line segment extraction. This approach reduces a 3D line segment fitting problem into two 2D line segment fitting problems and takes advantage of both images and depth maps. In our method, 3D line segments are fitted incrementally along detected edge segments via minimizing fitting errors on two planes. By clustering the detected line segments, the resulting 3D representation of the scene achieves a good balance between compactness and completeness. Our experimental results show that the 3D line segments generated by our method are highly accurate. As an application, we demonstrate that these line segments greatly improve the quality of 3D surface reconstruction compared to a feature point based baseline.

In this paper, we propose a novel real-time method for tracking planar edge templates. This method tracks an edge template by estimating its homography transformations with respect to the sampled edge pixels detected from the incoming frames. Particularly, we define a cost function based on a new feature map of the to-be-tracked edge template and optimize it by a Lucas-Kanade-like algorithm. The feature map is defined as the fourth root of the distance transform. Our method operates on just edges so that it is good at tracking those low textured targets, such as hollow targets (mug rim), thin targets (cable, ring) and non-Lambertian objects (disc). We validate and compare our method with four other methods on five newly collected real-world video sequences. The results achieves the lowest overall average error (1.58 pixels) and also outperforms others in terms of success rate. The per frame processing time of about 30 ms proves that our method is acceptable in realtime applications. The code and dataset are publicly available at: http://webdocs.cs.ualberta.ca/~xuebin/.

This paper presents a novel boundary based semiautomatic tool, ByLabel, for accurate image annotation. Given an image, ByLabel first detects its edge features and computes high quality boundary fragments. Current labeling tools require the human to accurately click on numerous boundary points. ByLabel simplifies this to just selecting among the boundary fragment proposals that ByLabel automatically generates. To evaluate the performance of By-Label, 10 volunteers, with no experiences of annotation, labeled both synthetic and real images. Compared to the commonly used tool LabelMe, ByLabel reduces image-clicks and time by 73% and 56% respectively, while improving the accuracy by 73% (from 1.1 pixel average boundary error to 0.3 pixel). The results show that our ByLabel outperforms the state-of-the-art annotation tool in terms of efficiency, accuracy and user experience. The tool is publicly available: http://webdocs.cs.ualberta.ca/~vis/ bylabel/.