Search

Intelligent bus system plays a key role in the modern smart city. The number of passengers in the buses or at the stations is necessary for making an optimal scheduling policy of public buses. We develop a crowd counting algorithm to provide the counting information for a bus dispatch system in a human–machine system. In consideration of the challenges (e.g., pedestrian occlusions, non-uniform crowd distributions, and scale variations) existed in hand-crafted features based crowd counting, a scale-distribution-aware multi-column convolutional neural network (SDA-MCNN) is presented to count crowds by summing up the output (denoted as the density map) of the SDA-MCNN. The SDA-MCNN is robust to scale variations by processing a crowd image with multiple convolutional neural network (CNN) columns and minimizing the per-scale loss. A weighted Euclidean loss is proposed to handle non-uniform crowd distributions. The loss can increase activations in dense regions and restrain activations in backgrounds. A new approach to estimate perspective maps of dense crowds is put forward to offer necessary information for generating density maps with human-shaped kernels. Evaluations on benchmarks are performed with other state-of-the-art counting approaches using deep neural networks. Comparative results verify the accuracy of our counting approach in challenging crowds. Evaluations on the real world BUS data reveal the accuracy of the proposed approach in counting passengers in spite of the complex environment.

Complexity is a major obstacle to apply formal verification techniques in industry. Assume-guarantee (AG) reasoning is a promising compositional verification approach to deal with complexity. Proper decomposition and assumption design is vital in AG reasoning. We are motivated by the fact that a good insight into the target system can help the formal verification. So, instead of trying to automate the AG reasoning process, an accident model called STAMP is adopted to help us get a better understanding of the safety control system to be verified and a strategy for decomposition and designing assumptions accordingly. This conceptual approach is exemplified by verifying a simplified train control system.

Crowd counting from low-resolution images is a challenging task, in particular in the edge computing system. An embedded equipment is commonly incompetent at patch-based crowd counting with real-time performance. This work develops a real-time method to count passengers in a bus by using Nvidia TX2. The videos of entry are recorded by a camera up ahead, and the data suffer from severe occlusion, which makes designing handcrafted features difficult. The counting is performed by summing up pixel values of the density map estimated using a compact convolutional neural network (CCNN), which is robust to scale variations by employing skip connections. A weighted Euclidean loss is proposed to handle cluttered backgrounds and blurry foregrounds. The loss increases the activations in dense regions, but can restrain the activations in background regions. The counting results are further improved by smoothing, which utilizes constraints between consecutive frames. Comparisons with existing counting approaches, including patch-based and whole image-based approaches, are made on two benchmarking datasets. The results indicate the accuracy of CCNN in counting dense crowds. Moreover, the evaluated bus datasets verify the feasibility of CCNN in counting passengers from low-resolution input images with real-time performance on TX2.

An attitude algorithm is used by a given magnetic and inertial measurement unit (MIMU) for weight fusion of the outputs of gyroscope and the outputs of the subunit constructed by accelerometer and magnetometer for attitude calculation of a rigid body. It has been proved that the weight assigned to the outputs of gyroscope determines the overall performances of attitude algorithm. A new weight adaptive regulation method based on Kalman filter and hidden Markov model is proposed, and then a new attitude algorithm is constructed by the integration of that method with recursive quaternion estimation algorithm. The authors’ attitude algorithm is compared with some other related attitude algorithms in their experiments. With the outputs of the sensors in MTi as the inputs of the new attitude algorithm, the maximum attitude estimation errors in response to three-dimensional random movements and the changes of their amplitudes are 2.85° and 6.82°, respectively, and the Allan variance of attitude drift in response to motionless conditions is 2.1×10−4°. Other attitude algorithms cannot simultaneously achieve similar dynamic and similar static performances comparing to their attitude algorithm. The results validate the overall performance improvement of their attitude algorithm.

Magnetic and inertial measurement units (MIMUs) are promising tools for attitude tracking of moving bodies without location restriction. An extended Kalman filter (EKF) is a commonly used attitude algorithm for MIMUs, and its Kalman gain is usually regulated according to the measurements of the accelerometer for the best integrated performance, i.e., the best performance for both when the carrier is motionless and when the carrier is moving. A hidden Markov model (HMM) is introduced, and then trained using static measurements of the accelerometer. Once the body has a movement, the match probability between the dynamic measurements of the accelerometer and the trained HMM will decrease, which is then used for the timely regulation of the Kalman gain to make the EKF rely more on the measurements of the gyroscope for attitude calculation. A slight revision to the introduced HMM is given for the improvement of the smoothness of the outputs of the HMM when the carrier is motionless. A relationship between the Kalman gain and the output of the HMM is also given, and the value range of the outputs of the HMM is readjusted in order to fit that relationship. Six competitive methods are compared with our method, and simulation and experiment tests validate the superiority of our method.

Communication-Based Train Control (CBTC) systems have been increasingly implemented on metro systems because of their characteristics, which result in safety and capacity improvements for metro operations. Automatic testing methods such as Model-Based Testing (MBT) have been applied to solve some particularly simple and ideal case studies. To bring automation to HIL testing, the authors apply MBT to HIL testing and present a novel hybrid online MBT testing platform that combines formal modelling methods with simulation. The theoretical methodology of the hybrid MBT, the platform architecture, and the testing results produced by the platform are described with a case study of a system under test (SUT) of a real Vehicle On-Board Controller (VOBC).

Aiming at the location problem of high-pressure pipeline leakage, this paper proposes a method for predicting the disturbed position of a Sagnac distributed optical fiber sensing system based on ensemble learning of convolutional neural networks (CNNs). A Stacking ensemble method is used to combine two different CNN models to improve the stability and location accuracy of the disturbance position prediction model. By training the prediction model using the spectrum features of interference signals generated by the disturbances at partial sensing positions, accurate prediction of an arbitrary disturbance position can be realized. The disturbance position prediction was carried out numerically on a sensing fiber with an effective length of 8.5 km. Simulation results show that a mean absolute error of no more than 14.6 m and a location resolution of 10 m are achieved. The method is insensitive to noise, low in system complexity, simple in data processing and accurate in location results.

Train trajectory optimisation plays a key role in improving energy saving performance and it is currently receiving increasing attention in railway research because of rising energy prices and environmental concerns. There have been many studies looking for optimal train trajectories with various different approaches. However, very few of the results have been evaluated and tested in practice. This study presents a field test of an optimal train trajectory on a metro line to evaluate the performance and the practicability of the trajectory with respect to operational energy computation. A train trajectory optimisation algorithm has been developed specifically for this purpose, and a field test of the obtained trajectory has been carried out on a metro line. In the field test, the driver controls the train in accordance with the information given by a driving advisory system, which contains the results of the train trajectory optimisation. The field test results show that, by implementing the optimal train trajectory, the actual energy consumption of the train can be significantly reduced, thereby improving the operational performance. Moreover, the field test results are very similar to the simulation results, proving that the developed train kinematics model is effective and accurate.

In this paper, a method of fault diagnosis for the on-board equipment based on Timed Automata (TA) and mutation theory is proposed. Firstly, several kinds of mutation operators were injected in TA model to get Fault Timed Automta (FTA), in which the fault with timed features is described. Secondly, the approach of fault diagnosis and the verification of the diagnosability were proposed by parallel product A, in which the necessary and sufficient conditions of fault diagnosiability were discussed. Finally, we take the Ceiling Speed Monitoring (CSM) and Target Speed Monitoring (TSM) functions as a case study to present the feasibility of the method. Based on the CSM diagnoser, the fault set with events and time behaviors was effectively divided into two categories: diagnosable and non-diagnosable. The result is shown that the fault diagnosis and verification of diagnosability can be performed and it is of great significance for fault diagnosis of the on-board equipment.

Maximizing regenerative energy utilization is an important way to reduce energy consumption in subway systems. Regenerative energy can be immediately utilized by traction trains, or be stored in energy storage systems and reused later. This paper adopts a combination of both methods, and proposes a dual-objective integrated optimization problem to simultaneously minimize the total net traction energy consumption and the investment of energy storage systems. Both timetable and configuration of energy storage systems are controlled in the optimization problem. A mathematical model of the dual-objective optimization problem is first formulated. Then to solve it, an c-constraint method is designed to transform the dual-objective optimization problem into several single-objective optimization problems, and an improved artificial bee colony algorithm is designed to solve them sequentially. Finally, numerical examples based on the actual data from a subway system in China are performed to show the effectiveness of the proposed method. Experimental results indicate that net traction energy consumption is effectively reduced by applying the proposed method, and it reduces to different degrees when using different size of energy storage systems, with an correspondingly optimized timetable. A set of Pareto optimal solutions is obtained for the experimental subway system, which is helpful for decision makers.

The communications-based train control (CBTC) is a typical safety-critical system that protects and directs train operations in urban rail transit. It is suggested to provide on-going safety protections for the automatic train protection, which is a kernel function of the CBTC. Runtime verification is a technique for monitoring system executions against safety requirements. A particular challenge in implementing of a runtime verification system for the CBTC is the appropriate monitor specification. This paper presents a novel dynamic monitoring generation method to the problem. The train control procedures of the CBTC is formalized by parametric hybrid automata (PHA), which introduces notations of parametric expressions for flow, transition conditions and invariants. With an observation, the PHA is instantiated to a standard hybrid automaton. The monitor specification is then generated automatically by calculating the reachable set of the automaton with respect to some selected safety-related properties. The presented method is evaluated in a hard-ware in the loop CBTC platform, which is developed with realistic engineering data of Beijing Yizhuang metro line. The experiment results show that the approach is feasible, and various dangerous of the CBTC system are prevented from developing into accidents of train collisions.

As an one-time braking mode curve, which is generated by the Target Speed Monitor (TSM) in Chinese Train Control System Level 3 (CTCS -3) based on the target speed, target distance, line condition, and train characteristics, performs the core safety critical functions. However, it is characterized by large parameters input domain, complex fault modes, and complicated combinations of various parameters and fault modes, which make it difficult to be effectively and efficiently tested in the train on-board equipment. In this paper, a methodology of complete test suites generated technology for the CTCS -3 TSM has been presented based on combinatorial testing (CT) strategies theory. Firstly, according to the system requirement specification (SRS) of CTCS -3, the combinatorial testing model of CTCS -3 TSM is built, which considering the selection of system parameters and its related constraints. Secondly, one of the combination strategies, in parameter order general (IPOG), is applied to obtain the test suites under different coverage criteria. Finally, the mutation analysis has been introduced to detect the ability of the generated test suites with different faults. We also show some interesting results to compare with the test suites generated by the random testing in the same conditions.

A novel drive train system for pure electric vehicles is proposed. The system uses dual motor to achieve power-shift, and its transmission, which only uses one slide sleeve to shift gear uninterruptedly due to using one-way clutch, is based on parallel shaft gears and planetary gears. The structure and principle of the drive train is introduced. After the speed/torque relationship between power input and output is realized, downshift process can be described in detail. Accurate transmission model have been built and the dynamic control by PID controller of the downshift process is simulated. The simulation results show that the device has good performance in gear-shifting quality.

Extended Kalman Filter (EKF) and Complementary Filter (CF) are the two most commonly-used attitude determination algorithms in inertial and magnetic measurement units. It is known that the only difference between the a posterior attitude estimates provided by EKF and CF respectively is the gain matrix (GM) assigned to innovation for attitude update. Through mathematical derivation, it is concluded that the GM of EKF can be simplified to the one of CF, which means that CF is an approximation to EKF. Monte Carlo simulations were done to validate what influence of these simplifications is put on the performance of EKF. The main finding is EKF is more stable than CF in condition of low sampling rate, and hence CF must rely more on the measurements of gyroscope for attitude determination to improve its stability.

The distance-to-go curve, which performs the safety critical functions of the Chinese Train Control System Level 3 (CTCS-3), is characterized by large input domain parameters and complex fault modes. It is very difficult to completely test the distance-to-go curve of train onboard equipment. In this paper, a method of generating complete test suite for the CTCS-3 Target Speed Monitor (TSM) has been presented, based on Input Equivalence Class Partitioning (IECP) test theory. Firstly, according to the system requirement specification (SRS) of CTCS-3, the finite state machine models of TSM are built, which satisfy the different brake control modes of driver with high Priority and Automatic Train Protection (ATP) with high Priority. Secondly, the Input/Output (I/O) equivalence and the equivalence class partition theory are applied to obtain two kinds of IECPs with different braking control modes. Finally, the fault models and fault domain of system under test (SUT) are introduced. Two complete test suites of TSM are generated, which can detect any faults in its given fault domain. Different brake control modes in TSM have been compared, in which the ATP with high Priority control mode has a higher test cases number, but more time and memory consumption.

As a safety-critical system, the Chinese Train Control System Level 3 (CTCS-3) have complex fault modes. To generate test suite, which can cover all known faults, is very difficult. In this paper, we proposed a methodology of generating test suite for CTCS-3 train control system based on timed automata with input and output (TAIO) and mutation testing theory. Firstly, according to the characteristics of the fault modes in CTCS-3, mutation operators that contain all known faults of mode transition function (“change action”, “change target”, “change source”, etc) are designed. Secondly, the TAIO model of mode transition (TAIO A) in CTCS-3 is established. With each mutation operator, mutants of mode transition timed automata model (TAIO M) are built. Test suites are generated based on the conformance relation of TAIO A and TAIO M using the k-Bounded model checking technique. Finally, the coverage of the test suites have been analyzed by the conformance relation score (CRS), average conformance relation score (ACRS) and weighted conformance relation score (WCRS). Three fault modes of “change action”, “change invariant” and “sink location” could be covered effectively and completely detected, wherever, two fault modes coverage of “negate guard” and “invert reset” are not high enough, which need additional observation to detect.

To acquire the fruit region of green apple image in close color more completely, this work designed a method of region extraction before fruit region combination. Firstly, channel images in every color space were analyzed to determine appropriate color channel images: images of G and b color channel in RGB and Lab color spaces, respectively. Then, these images were segmented by K-means clustering to obtain segmented images of G and b color channel. Post-processing was conducted to different segmentation results. For example, open operation and micro-region elimination were conducted to segmented image of G color channel. The same processing was conducted to b color channel image after combining with above processed G color channel image and filling holes. Therefore, the target fruit region can be acquired completely. The validity of this method was verified by experiment.

Peach fruit has inhomogeneous colors including green, white and red. The work aimed at studying peach fruit recognition method under natural environment. Firstly, peach was conducted with image segmentation pretreatment using OTSU dynamic threshold segmentation method based on R-G color description. Fruit region was demarcated in original image. Secondly, peach fruit was distinguished using improved NNPROD algorithm after accelerated optimization. At last, the effectiveness of the method was evaluated by experiments.

Branches recognition of apple harvesting robot were studied to provide rich environment information for robot's harvesting obstacle avoidance. Firstly, the leaves and other background in the apple image were removed based on R-G color difference. Secondly, image segmentation was conducted using K-Means clustering based on “a” color feature in the Lab color space. Then the noise removal and image perfection operation were executed, and the central axis — the main characteristics of branches — was extracted through skeletonizing method and removing subbranch method. Finally, experiments were carried out on recognition of apple branches, indicating feasibility and effectiveness of the proposed method.

Magnetic and inertial measurement units (MIMU) are currently being explored as a promising tool for attitude tracking of a moving object, such as human body parts. The function of attitude calculation is realized by using attitude algorithms. The overall performance of these algorithms is seriously influenced by linear acceleration of the moving object. Therefore, there is a requirement to find solutions to this problem. In this paper, a new attitude algorithm for MIMU known as REQUEST is introduced. The algorithm is then revised in order to be suitable for MIMU. A new representation of linear acceleration of the moving object is then constructed. An artificial neural network (ANN) is used to establish the functional relationship between this representation and the weights assigned to the vector measurements in REQUEST, for the purpose of adaptively regulating the vector measurements according to the representation. In this way, the measurements of the gyroscope can be relied on more for attitude calculation when the carrier has a higher linear acceleration. A Kalman filter (KF) is also used prior to the establishment of the functional relationship in order to take full advantage of historical sensor information for accurate estimation of the representation. Our experiments have verified good static and dynamic performances of our KF+ANN-based REQUEST algorithm.