Design and Experimental Evaluation of an Aerial Solution for Visual Inspection of Tunnel-like Infrastructures

Current railway tunnel inspections rely on expert operators performing a visual examination of the entire infrastructure and manually annotating encountered defects. Automatizing the inspection and maintenance task of such critical and aging infrastructures has the potential to decrease the associated costs and risks. Contributing to this aim, the present work describes an aerial robotic solution designed to perform autonomous inspections of tunnel-like infrastructures. The proposed robotic system is equipped with visual and thermal sensors and uses an inspection-driven path planning algorithm to generate a path that maximizes the quality of the gathered data in terms of photogrammetry goals while optimizing the surface coverage and the total trajectory length. The performance of the planning algorithm is demonstrated in simulation against state-of-the-art methods and a wall-following inspection trajectory. Results of a real inspection test conducted in a railway tunnel are also presented, validating the whole system operation.

Results of laboratory and bench-scale studies of seed dispensers with an electric drive of working tools

Mechanical and pneumatic seed drills of both domestic and foreign production are used in Russian farms. They are equipped with a mechanical drive of working tools and an electronic seeding control system. Due to the slipping of the wheels or the breakage of the chains, the sowing of seeds in individual seed dispensers interrupts. According to the results of laboratory and bench-scale studies in respect to soybean seeds, the required power for the electric drive of one seed dispenser was determined, which, depending on the disk rotation speed from 10 to 60 rpm, ranged from 30 to 120 W. By calculation, using the analytical expression, the power, required for the fan drive of a 12-row seed drill, was determined, which, depending on the disk rotation speed, ranged from 1.6 to 2.47 kW. A condition is formulated, which will eliminate the probability of shifting and rolling seeds along the furrow after their fall out of the sowing disc rotating in the opposite direction to the movement of the seeder unit, provided correspondence of the linear speed of the sowing disc and the speed of the seeder unit (the effect of zero overlaps). In this case, the trajectory length of the seeds falling to the furrow should be consistent with the speed of the seeder unit and the seeding rate according to the proposed expression.

Converting POMDPs into MDPs using History Representation

Defining utility functions like the V and Q functions is essential for developing reinforcement learning (RL) solutions for POMDPs. Ideally, we want to define these functions over histories of observations and actions, which the agent can observe. However, the number of possible histories grows exponentially with the trajectory length, making it impractical to reliably estimate history-based utility functions. A common solution is to construct a compact representation of the history. Recently, with the resurgence of deep learning, researchers use recurrent neural networks to learn low-dimensional history representations. Conversion from POMDPs to MDPs using history representation appears to be so effective and seamless that it is common to see a theory-practice gap in deep RL papers, where an algorithm is theoretically formulated in an MDP setting but is empirically evaluated on POMDP tasks without any justifications on why it would work in the latter setting. Thisdocument provides a justification for the conversion from POMDPs to MDPs using history representation.

Recovering mixtures of fast diffusing states from short single particle trajectories

Single particle tracking (SPT) directly measures the dynamics of proteins in living cells and is a powerful tool to dissect molecular mechanisms of cellular regulation. Interpretation of SPT with fast-diffusing proteins in mammalian cells, however, is complicated by technical limitations imposed by fast image acquisition. These limitations include short trajectory length due to photobleaching and shallow depth of field, high localization error due to the low photon budget imposed by short integration times, and cell-to-cell variability. To address these issues, we developed methods to infer distributions of diffusion coefficients from SPT data with short trajectories, variable localization accuracy, and absence of prior knowledge about the number of underlying states. We discuss advantages and disadvantages of these approaches relative to other frameworks for SPT analysis.

Trajectory planning in UAV emergency networks with potential underlaying D2D communication based on K-means

At present, unmanned aerial vehicles (UAVs) have been widely used in communication systems, and the fifth-generation wireless system (5G) has further promoted the vigorous development of them. The trajectory planning of UAV is an important factor that affects the timeliness and completion of missions, especially in scenarios such as emergency communications and post-disaster rescue. In this paper, we consider an emergency communication network where a UAV aims to achieve complete coverage of potential underlaying device-to-device (D2D) users. Trajectory planning issues are grouped into clustering and supplementary phases for optimization. Aiming at trajectory length and sum throughput, two trajectory planning algorithms based on K-means are proposed, respectively. In addition, in order to balance sum throughput with trajectory length, we present a joint evaluation index. Then relying on this index, a third trajectory optimization algorithm is further proposed. Simulation results show the validity of the proposed algorithms which have advantages over the well-known benchmark scheme in terms of trajectory length and sum throughput.

Second sacral sacralalar‐iliac (S2AI) screw placement in adult degenerative scoliosis (ADS) patients: an imaging study

Background The imaging characteristics of sacral sacralalar-iliac (S2AI) screw trajectory in adult degenerative scoliosis (ADS) patients will be determined. Methods S2AI screw trajectories were mapped on three-dimensional computed tomography (3DCT) reconstructions of 40 ADS patients. The starting point, placement plane, screw template, and a circle centered at the lowest point of the ilium inner cortex were set on these images. A tangent line from the starting point to the outer diameter of the circle was selected as the axis of the screw trajectory. The related parameters in different populations were analyzed and compared. Results The trajectory length of S2AI screws in ADS patients was 12.00 ± 0.99 cm, the lateral angle was 41.24 ± 3.92°, the caudal angle was 27.73 ± 6.45°, the distance from the axis of the screw trajectory to the iliosciatic notch was 1.05 ± 0.81 cm, the distance from the axis of the screw trajectory to the upper edge of the acetabulum was 1.85 ± 0.33 cm, and the iliac width was 2.12 ± 1.65 cm. Compared with females, the lateral angle of male ADS patients was decreased, but the trajectory length was increased (P < 0.05). Compared to patients without ADS in previous studies, the lateral angle of male patients was larger, the lateral angle of female patients was increased, and the caudal angle was decreased (P < 0.05). Conclusions There is an ideal trajectory of S2AI screws in ADS patients. A different direction should be noticed in the placement of S2AI screws, especially in female patients.

Feasibility, safety and accuracy of a CT-guided robotic assistance for percutaneous needle placement in a swine liver model

Evaluate the feasibility, safety and accuracy of a CT-guided robotic assistance for percutaneous needle placement in the liver. Sixty-six fiducials were surgically inserted into the liver of ten swine and used as targets for needle insertions. All CT-scan acquisitions and robotically-assisted needle insertions were coordinated with breath motion using respiratory monitoring. Skin entry and target points were defined on planning CT-scan. Then, robotically-assisted insertions of 17G needles were performed either by experienced interventional radiologists or by a novice. Post-needle insertion CT-scans were acquired to assess accuracy (3D deviation, ie. distance from needle tip to predefined target) and safety. All needle insertions (43/43; median trajectory length = 83 mm (interquartile range [IQR] 72–105 mm) could be performed in one (n = 36) or two (n = 7) attempts (100% feasibility). Blinded evaluation showed an accuracy of 3.5 ± 1.3 mm. Accuracy did not differ between novice and experienced operators (3.7 ± 1.3 versus 3.4 ± 1.2 mm, P = 0.44). Neither trajectory angulation nor trajectory length significantly impacted accuracy. No complications were encountered. Needle insertion using the robotic device was shown feasible, safe and accurate in a swine liver model. Accuracy was influenced neither by the trajectory length nor by trajectory angulations nor by operator’s experience. A prospective human clinical trial is recruiting.

Trajectory Planning in UAV Emergency Networks with Potential Underlaying D2D Communication Based on K-means

Unmanned aerial vehicles (UAVs) have been widely used in communication systems due to excellent maneuverability and mobility. The ultra-high speed, ultra-low latency, and ultra-high reliability of 5th generation wireless systems (5G) have further promoted vigorous development of UAVs. Compared with traditional means of communication, UAV can provide services for ground terminal without time and space constraints, so it is often used as air base station (BS). Especially in emergency communications and rescue, it provides temporary communication signal coverage service for disaster areas. In the face of large-scale and scattered user coverage tasks, UAV's trajectory is an important factor affecting its energy consumption and communication performance. In this paper, we consider a UAV emergency communication network where UAV aims to achieve complete coverage of potential underlying D2D users (DUs). The trajectory planning problem is transformed into the deployment and connection problem of stop points (SPs). Aiming at trajectory length and sum throughput, two trajectory planning algorithms based on K-means are proposed. Due to the non-convexity of sum throughput optimization, we present a sub-optimal solution by using the successive convex approximation (SCA) method. In order to balance the relationship between trajectory length and sum throughput, we propose a joint evaluation index which is used as an objective function to further optimize trajectory. Simulation results show the validity of the proposed algorithms which have advantages over the well-known benchmark scheme in terms of trajectory length and sum throughput.

Traffic Network Identification Using Trajectory Intersection Clustering

The current airspace route system consists mainly of pre-defined routes with a low number of intersections to facilitate air traffic controllers to oversee the traffic. Our aim is a method to create an artificial and reliable route network based on planned or as-flown trajectories. The application possibilities of the resulting network are manifold, reaching from the assessment of new air traffic management (ATM) strategies or historical data to a basis for simulation systems. Trajectories are defined as sequences of common points at intersections with other trajectories. All common points of a traffic sample are clustered, and, after further optimization, the cluster centers are used as nodes in the new main-flow network. To build almost-realistic flight trajectories based on this network, additional parameters such as speed and altitude are added to the nodes and the possibility to take detours into account to avoid congested areas is introduced. As optimization criteria, the trajectory length and the structural complexity of the main-flow system are used. Based on these criteria, we develop a new cost function for the optimization process. In addition, we show how different traffic situations are covered by the network. To illustrate the capabilities of our approach, traffic is exemplarily divided into separate classes and class-dependent parameters are assigned. Applied to two real traffic scenarios, the approach was able to emulate the underlying route systems with a difference in median trajectory length of 0.2%, resp. 0.5% compared to the original routes.

Crowd-Steer: Realtime Smooth and Collision-Free Robot Navigation in Densely Crowded Scenarios Trained using High-Fidelity Simulation

We present a novel high fidelity 3-D simulator that significantly reduces the sim-to-real gap for collision avoidance in dense crowds using Deep Reinforcement Learning (DRL). Our simulator models realistic crowd and pedestrian behaviors, along with friction, sensor noise and delays in the simulated robot model. We also describe a technique to incrementally control the randomness and complexity of training scenarios to achieve better convergence and generalization capabilities. We demonstrate the effectiveness of our simulator by training a policy that fuses data from multiple perception sensors such as a 2-D lidar and a depth camera to detect pedestrians and computes smooth, collision-free velocities. Our novel reward function and multi-sensor formulation results in smooth and unobtrusive navigation. We have evaluated the learned policy on two differential drive robots and evaluate its performance in new dense crowd scenarios, narrow corridors, T and L-junctions, etc. We observe that our algorithm outperforms prior dynamic navigation techniques in terms of metrics such as success rate, trajectory length, mean time to goal, and smoothness.