Over the past few years, self-supervised monocular depth estimation has received widespread attention. Most efforts focus on designing different types of network architectures and loss functions or handling edge cases, for example, occlusion and dynamic objects. In this work, we take another path and propose a novel conditional diffusion-based generative framework for self-supervised monocular depth estimation, dubbed MonoDiffusion. Because the depth ground-truth is unavailable in a self-supervised setting, we develop a new pseudo ground-truth diffusion process to assist the diffusion for training. Instead of diffusing at a fixed high resolution, we perform diffusion in a coarse-to-fine manner that allows for faster inference time without sacrificing accuracy or even better accuracy. Furthermore, we develop a simple yet effective contrastive depth reconstruction mechanism to enhance the denoising ability of model. It is worth noting that the proposed MonoDiffusion has the property of naturally acquiring the depth uncertainty that is essential to be implemented in safety-critical cases. Extensive experiments on the KITTI, Make3D and DIML datasets indicate that our MonoDiffusion outperforms prior state-of-the-art self-supervised competitors. The source code will be publicly available upon the acceptance.

Over the past few years, monocular depth estimation and completion have been paid more and more attention from the computer vision community because of their widespread applications. In this paper, we introduce novel physics (geometry)-driven deep learning frameworks for these two tasks by assuming that 3D scenes are constituted with piece-wise planes. Instead of directly estimating the depth map or completing the sparse depth map, we propose to estimate the surface normal and plane-to-origin distance maps or complete the sparse surface normal and distance maps as intermediate outputs. To this end, we develop a normal-distance head that outputs pixel-level surface normal and distance. Afterthat, the surface normal and distance maps are regularized by a developed plane-aware consistency constraint, which are then transformed into depth maps. Furthermore, we integrate an additional depth head to strengthen the robustness of the proposed frameworks. Extensive experiments on the NYU-Depth-v2, KITTI and SUN RGB-D datasets demonstrate that our method exceeds in performance prior state-of-the-art monocular depth estimation and completion competitors.

We propose a novel reactive algorithm for a multi-robot formation navigating from its starting point to its destination in geometrically changing non-convex environments containing dynamic obstacles that move stochastically. Taking formation’s buffer into consideration, the proposed algorithm estimates the optimal velocity by means of minimizing collision threats from obstacles to the formation. It considers the obstacles in pairs so that the formation crosses the gap between every involved pair of obstacles. At every time step, this algorithm is executed in four phases: 1) optimal velocity estimation using a probabilistic model; 2) topological trajectory decision; 3) optimal waypoint (crossing point) determination; and 4) real-time optimal velocity determination. Compared to existing algorithms, the simulation shows that our algorithm is robust and has significant improvement.Note to Practitioners—This paper was motivated by the problem of formation navigation in places containing a large number of moving objects such as human beings and other vehicles. Existing approaches either do not forecast the probability of collision, or do not explicitly consider a certain comfort or safety distance to maintain. This paper quantifies the expectation of objects getting “too close” to the formation before collision as a numerical collision forecast, where the future motions of objects are unknown: this allows a formation to navigate by acquiring least environment information. From this expectation, by considering the formation to be moving between pairs of these objects, the optimal path to the destination for the formation is computed. Our approach is proved by the simulation to have significant improvement compared to the existing approaches. In future research, we will consider more ways to avoid obstacles such as formation rotation and waiting.

Existing methods detect the keypoints in a non-differentiable way, therefore they can not directly optimize the position of keypoints through back-propagation. To address this issue, we present a partially differentiable keypoint detection module, which outputs accurate sub-pixel keypoints. The reprojection loss is then proposed to directly optimize these sub-pixel keypoints, and the dispersity peak loss is presented for accurate keypoints regularization. We also extract the descriptors in a sub-pixel way, and they are trained with the stable neural reprojection error loss. Moreover, a lightweight network is designed for keypoint detection and descriptor extraction, which can run at 95 frames per second for 640x480 images on a commercial GPU. On homography estimation, camera pose estimation, and visual (re-)localization tasks, the proposed method achieves equivalent performance with the state-of-the-art approaches, while greatly reduces the inference time.

The concept of stimuli-induced equilibrium point (SIEP) has been recently introduced to characterize the psychological interaction of a ramp driver with its putative leader and follower on the expressway during a merging maneuver. It enables the computation of the reference target gap speed and position for the on-ramp merging vehicle based on current traffic conditions and ramp vehicle response. The SIEP has been shown to improve the performance of existing automatic ramp merging control strategies while increasing the level of safety during the merging maneuver. The performance and advantages of the SIEP-based approach have been assessed only through numerical simulations. In this paper, we conducted a comprehensive experimental evaluation of the performance and safety of this SIEP-based approach to ramp merging control using a lab-based test-bed. We employed a novel Pc metric, which is based on the concept of probability of collision, to perform a systematic validation of the SIEP safety. Such a metric serves as a standardized methodology to quantitatively compare the SIEP-based approach with those in this paper.

Classical convolutional neural networks used in computer vision tasks perform excellently in accuracy, but they are unsatisfactory in computational cost especially with the networks going deeper and the image size going larger. Special models based on visual attention have showed their advantages in dealing with spatial information for saving computational cost at inference time. These models are designed to imitate human visual attention mechanism, but they are not able to achieve realize adaptive receptive scope for different object size. In this paper, a recurrent location and scope selection approach is proposed to improve the attention efficiency, which is more similar to human visual mechanism. We evaluate our model on the basic visual recognition task, where it outperforms the baselines and could provide approximated bounding boxes in a weakly supervised way.

In this paper, we propose the novel concept of stimuli-induced equilibrium point to synthesize the speed and position references for automatic on-ramp merging systems. Based on the psychological field theory, the intensity of the stimuli that act upon a driver between two vehicles in a three-vehicle platooning configuration is mathematically modeled to calculate a point at which the stimuli resultant becomes zero. This approach intends to mimic drivers' decision process when certain distance separation with respect to the leader and follower vehicles is attained for safety. The location of this point is continuously updated according to the speed of the middle vehicle and the current traffic scenario. Such stimuli-induced equilibrium point has shown to improve the performance of existing automatic merging control schemes while increasing safety conditions by providing enough reaction time for drivers to avoid an eventual collision.

Rigid actuators that accompany traditional robotic systems allow them to perform tasks with unmatched precision and efficiency. However, such systems are inapt when flexibility and suppleness are required. Inversely, soft robotic actuators are constructed out of naturally compliant materials. This compliance allows them to execute complex motions while their inherent lack of hardness removes the dangers that commonly accompany traditional robots. One archetype of the soft robot is the pneumatic bending actuator. The actuators convert the flow of pressurized air into bending outputs; however, most bending actuators are only limited to one bending trajectory. A new bending actuator, capable of active reconfiguration, is realized through the combination of a non-inflating, flexible, plastic spine and multiple, inflatable, fabric modules. The spine acts as the primary conduit and is responsible for distributing the pressurized flow to the modules. Modules can be removed and replaced seamlessly. The combination of varying module geometries can produce a heterogenic bending trajectory. Blockages located along the spine can separate flow, allowing more customization of the trajectory. A physical model predicting the pressure-bending relationship was conceptualized and verified. When pressurized to 100kPa, the 5-module and 10-module actuators achieved bending angles of 48.76° and 137.65° respectively. The actuators behave as a first-order system when supplied with a step input.

In this brief, we propose a dynamic surface control with actuator failure compensation for a class of feedback linearizable systems with locally Lipschitz nonlinearities. First, a dynamic surface state feedback control scheme is designed, which incorporates radial basis function networks in a novel approach, to compensate system uncertainties and dynamic changes induced by actuator failures. Then, an output feedback controller is obtained by means of high-gain observers. It is proved that our control schemes guarantee the uniform ultimate boundedness of the system, and that the output tracking error converges to an arbitrarily small residual set. Finally, a simulation is carried out to illustrate the performance of the designed control schemes.

In this paper, we report our investigation on the force-induced responses of zebrafish chorion before and after being perturbed by an externally applied prescribed force. Our experimental results show that application of an external force leads to a reduction in the stiffness of the zebrafish chorion, and that variation in such a reduction is strongly influenced by the perturbation time. Our results provide evidence supporting the hypothesis that the stiffness of some cellular organism may be modified by the controlled application of an appropriate mechanical force, and demonstrate the potential of force-feedback control in micro/nano-manipulation as a promising engineering technique for studying disease properties and manipulating cellular biological functions.

This paper presents an approach to optimize the zebrafish-embryo microinjection process by selecting suitable speed trajectories for the motion of the micropipette, with the objective of minimizing the risk of damaging the embryo during the microinjection process. A viscoelastic model of a zebrafish embryo is developed based on experimental data. The speed optimization problem is formulated based on this viscoelastic model. Simulation results indicate that there exists an optimal speed trajectory for the micropipette under certain conditions. The key benefit of this speed optimization approach is that the turgor pressure inside the membrane is prevented from increasing significantly during the microinjection process, thus avoiding damaging the biological components in the embryo.

This is the first part of the whole paper and in this part, characteristics of two similar nonlinear feedback control laws of differential mobile robots were investigated in a comparative framework. First evolutions of robot's heading under those two controls are studied. Then uniqueness of robot's trajectory with respect to gain ratio and characteristics of trajectory curvature are revealed. For the first time, the concept of ldquocritical gain ratiordquo is proposed. Simulation results are also presented.

This is the second part of the whole paper and in this part, the robustness of a nonlinear control law for mobile robot is examined under Invariance Principles. It provides insight into the stable zone for a given set of controller gains and make possible a more robust controller under some circumstances. Improvement to this control law is achieved from the analysis results and is verified in Matlab simulation. Implementation with real robots has been done to demonstrate the application to multiple robot formation control.

In this paper, we investigate stability of a nonlinear tracking control for a mobile robot in connection with the method of artificial potential trench. We first refine the concept of potential trench function and propose several approaches to construct potential trench functions. Then we synthesize a control law that stabilizes a robot while tracking a goal point moving on a given segment (smooth curve) and extend it to a simple case of formation control. Results from computer simulation verify and demonstrate the effectiveness of this novel control.

Stability and performance degradation in feedback control systems under saturation can be overcome by anti-windup techniques. In this paper, a new anti-windup approach called “Auxiliary Input Anti-windup” is proposed, which incorporates an auxiliary input into the controller of the system. Finding the maximal output admissible set of the augmented closed-loop system, the control effort is then obtained from a quadratic optimization problem subjected to linear inequality constraints. Performance and effectiveness of the proposed anti-windup approach, as compared with other methods, is illustrated with simulation examples. Experimental results show that the introduced Auxiliary Input (AI) anti-windup scheme improves the performance of the system.

The microinjection is a common technique in genetic engineering for transferring genetic material into a cell. It is normally performed on a micromanipulation system that usually consists of an inverted microscope, a micromanipulator, a micropipette and an injector. Manual microinjection is a conventional and widespread practice. However, this approach suffers from low success rate, poor efficiency, and high probability of contamination. We have developed a micromanipulation system for automatic batch microinjection by automating the key process of penetration in a microinjection. The microinjection system automatically identifies and penetrates each embryo/cell with a constant and fast speed, and in doing so, eliminates the possibility of contamination caused by manual operations. The effectiveness of this force-controlled system has been experimentally demonstrated in the tasks of automatic identification and penetration zebrafish embryos, which are widely used as a model for studying vertebrate development and genetics.

This paper presents a discrete-event approach to synthesis of transcription control for a class of (computational) gene networks. Given a set of genes and protein-gene and/or protein-protein interaction patterns, this approach can be used to synthesis a transcription control induced by these interaction patterns, and as a consequence, determine the behavior of the gene network. This proposed approach is demonstrated in the synthesis of a transcription control that produces the observed (logical) behavior in the induction of a phage ⁁ lysogen.

In micromanipulation, the size of the manipulated object is usually much less than 1 mm in a single dimension, in which case gravitational and inertial forces are no longer dominant. This leads to problems (for manipulation through force) that are not evident in the macroworld, and for which the macroworld techniques alone may not be adequate to provide solutions. This paper surveys critical issues and their available solutions related to force control in micromanipulation. It focuses on: 1) techniques for dealing with adhesion forces and 2) methods for force sensing and control

In this paper, a dynamic scheduling algorithm is developed to optimize the performance digital control systems by allocating CPU time efficiently. The periods of the control tasks are regarded as variables with dependency on the task current importance and scheduling constraints. The task current importance can be measured by the current cutoff frequency of the control signal. It indicates the task current requirement on sampling frequency. CPU time is allocated under the consideration of task current requirement and the performance index change-rate to optimize the overall system performance. The simulation results verify that this scheduling algorithm can improve the overall system performance under limited computing resource.

Automation of manufacturing workcell relies heavily on sensors and logic control systems to ensure smooth process flow through its devices. Commonly, such logic control systems (e.g. programmable logic controller) are setup for controlling workcells for a specific process flow. These systems require extensive restructuring of its control logic and re-wiring of the communication cables (e.g. feedback from sensors, relays to actuators) whenever there is need for the workcells to change their configuration for different process flow. These requirements result in loss of the production time and incur additional costs. In this paper a different approach, which has the potential of rapid configuration of logical control, is demonstrated on a modular robotic workcell. The approach uses computer network as the communication media and the control logic is developed using the supervisory control theory.

An approach to uncertainty compensation using a multilayer feedforward neural network in multi-manipulator system control is proposed. The proposed approach is developed by formulating the dynamics of the multi-manipulator system in the constrained motion framework. The error-backpropagation algorithm is employed for neural network learning. The teaching signal for neural network learning is derived by analyzing the stability of the closed-loop system. It is shown that if the neural network learns to generate the proper compensating signal, then the constrained motion of the multi-manipulator system tracks the desired motion asymptotically; as a consequence, the desired forces can be achieved. Computer simulations are conducted to verify the proposed approach.>

In this paper, the dynamic behavior of a three-layer feedforward neural network as a uncertainty compensator for robotic control is investigated. The investigation is conducted in the context of the robot trajectory tracking problem, where the neural network (with the error-backpropagation algorithm) is used as a uncertainty compensator in conjunction with the feedback linearization control (i.e. computed torque) and a PD control. Through computer simulation, it is verified that the dynamics of the neural network has a specific pattern when the learning rate is sufficiently small, and that such a specific pattern of weight variation in the neural network represents a sufficient condition for closed-loop system performance improvement.