Inverse Learning Control of Nonlinear Systems Using Support Vector Machines

An inverse learning control scheme using the support vector machine (SVM) for regression was proposed. The inverse learning approach is originally researched in the neural networks. Compared with neural networks, SVMs overcome the problems of local minimum and curse of dimensionality. Additionally, the good generalization performance of SVMs increases the robustness of control system. The method of designing SVM inverse learning controller was presented. The proposed method is demonstrated on tracking problems and the performance is satisfactory.     

The Cooperative Multi-agent Learning with Random Reward Values

This paper investigated how to learn the optimal action policies in cooperative multi-agent systems if the agents‘ rewards are random variables, and proposed a general two-stage learning algorithm for cooperative multiagent decision processes. The algorithm first calculates the averaged immediate rewards, and considers these learned rewards as the agents‘ immediate action rewards to learn the optimal action policies. It is proved that the learning algorithm can find the optimal policies in stochastic environment. Extending the algorithm to stochastic Markov decision processes was also discussed.     

Transit network design for small-medium size cities

This paper proposes a novel heuristic to solve the network design problem for public transport in small-medium size cities. Such cities can be defined as those with a diameter of a few kilometers with up to a few hundred thousand residents. These urban centers present a specific spatial configuration affecting the land use and mobility system. Transportation demand is widespread in origin and concentrated in a small number of attraction points close to each other. This particular structure of demand (‘many-to-few’) suggests the need for specific methodologies for the design of a transit system at a network level. In this paper, such design methodologies are defined in terms of models and solution procedures and tested on a selected case study. The solution methods show promising results. The key variables of the model are the routes and their frequencies. The constraints of the problem affect the overall demand to be served, the quality of the proposed service (transfer, load factors) and the definition of routes.     

基于相关性准则和R-ELM模型的岩溶隧道涌水量预测研究

为实现隧道涌水量的高精度预测,以相关系数法和极限学习机为理论基础,构建隧道涌水量预测模型。首先,结合工程实例对隧道涌水的影响因素进行分析,并利用相关系数法分析各因素与涌水量之间的相关性,以筛选出重要影响因素;其次,将筛选出的重要因素作为预测模型的输入层,并利用试算法和经验公式优化极限学习机的模型参数,再利用M估计弱化预测误差,进而构建出用于隧道涌水预测的R-ELM模型。研究表明： 1）岩溶隧道涌水灾害的影响因素较多,包括5类一级因素和12类二级因素,不同因素对隧道涌水灾害的影响程度存在一定差异; 2）R-ELM模型预测结果的平均相对误差仅为1.12%,具有较高的预测精度,不仅验证了模型参数优化和M估计优化的有效性,也验证了R-ELM模型在隧道涌水量预测中的适用性。     

柴油机双层隔振的自适应模糊控制方法及模拟试验的研究

该文针对船舶柴油机振动特点,提出了一种自适应模糊控制方法,构造了自适应模糊控制器,并可在线自适应调整模糊控制器的有关参数,由柴油机双层隔振系统模拟台架试验的结果表明,该模糊控制方法对柴油机的振动具有良好的控制效果。     

盲源分离技术在车辆轮胎噪声提取中的应用

介绍Kullback-Leibler发散度,推导基于该发散度和自然梯度算法的盲信号处理的计算方法,并通过一个仿真实例说明运用该方法对车辆轮胎噪音提取的可行性,从而为车辆的车轮故障诊断和车体音频信号的盲分离提供一个新的方法。     

Data mining in sloshing experiment database and application of neural network for extreme load prediction

Database of the sloshing model test has been mined. More than 540 terabytes experimental data have been accumulated for various cargo holds, vessels, environmental conditions, operational conditions, and experimental conditions. The database was organized, cleaned, and analyzed for the floating units larger than standard size LNG carriers or LNG fueled vessels. The selected target data was used for the machine learning to predict the model test results from the test conditions. An artificial neural network has been developed. Many different types of parameters were scaled and transformed as the input attributes followed by the optimization of the hyperparameters and the architecture. The network predicted the test results that were not used in the training process. The prediction results were validated according to the changes in the environmental conditions, operational conditions, and model dimensions. The accuracy of the network was acceptable to be applicable to the designing perspective.     

Position synchronization for track geometry inspection data via big-data fusion and incremental learning

Track geometry inspection data is important for managing railway infrastructure integrity and operational safety. In order to use track geometry inspection data, having accurate and reliable position information is a prerequisite. Due to various issues identified in this research, the positions of different track geometry inspections need to be aligned and synchronized to the same location before being used for track degradation modeling and maintenance planning. This is referred to as “position synchronization”, a long-standing important research problem in the area of track data analytics. With the aim of advancing the state of the art in research on this subject, we propose a novel approach to more accurately and expediently synchronize track geometry inspection positions via big-data fusion and incremental learning algorithms. Distinguishing it from other relevant studies in the literature, our proposed approach can simultaneously address data exceptions, channel offsets and local position offsets between any two inspections. To solve the Position Synchronization Model (PS-Model), an Incremental Learning Algorithm (IL-Algorithm) is developed to handle the “lack of memory” challenge for the fast computation of massive data. A case study is developed based on a dataset with data size of 18 GB, including 58 inspections between February 2014 and July 2016 over 323 km (200 miles) of tracks belonging to China High Speed Railways. The results show that our proposed model performs robustly against data exceptions via the use of multi-channel information fusion. Also, the position synchronization error using our proposed approach is within 0.15 meters (0.5 feet). Our proposed data-driven, incremental learning algorithm can quickly solve the complex, data-extensive, position synchronization problem, using an average of 0.1 s for processing one additional kilometer of track. In general, the data analysis methodology and algorithm presented in this paper are also suitable to address other relevant position synchronization problems in transportation engineering, especially when the dataset contains multiple channels of sensors and abnormal data outliers.     

Adaptive traffic signal control with actor-critic methods in a real-world traffic network with different traffic disruption events

The transportation demand is rapidly growing in metropolises, resulting in chronic traffic congestions in dense downtown areas. Adaptive traffic signal control as the principle part of intelligent transportation systems has a primary role to effectively reduce traffic congestion by making a real-time adaptation in response to the changing traffic network dynamics. Reinforcement learning (RL) is an effective approach in machine learning that has been applied for designing adaptive traffic signal controllers. One of the most efficient and robust type of RL algorithms are continuous state actor-critic algorithms that have the advantage of fast learning and the ability to generalize to new and unseen traffic conditions. These algorithms are utilized in this paper to design adaptive traffic signal controllers called actor-critic adaptive traffic signal controllers (A-CATs controllers).The contribution of the present work rests on the integration of three threads: (a) showing performance comparisons of both discrete and continuous A-CATs controllers in a traffic network with recurring congestion (24-h traffic demand) in the upper downtown core of Tehran city, (b) analyzing the effects of different traffic disruptions including opportunistic pedestrians crossing, parking lane, non-recurring congestion, and different levels of sensor noise on the performance of A-CATS controllers, and (c) comparing the performance of different function approximators (tile coding and radial basis function) on the learning of A-CATs controllers. To this end, first an agent-based traffic simulation of the study area is carried out. Then six different scenarios are conducted to find the best A-CATs controller that is robust enough against different traffic disruptions. We observe that the A-CATs controller based on radial basis function networks (RBF (5)) outperforms others. This controller is benchmarked against controllers of discrete state Q-learning, Bayesian Q-learning, fixed time and actuated controllers; and the results reveal that it consistently outperforms them.     

Driving analytics using smartphones: Algorithms,comparisons and challenges

The present work investigates the use of smartphones as an alternative to gather data for driving behavior analysis. The proposed approach incorporates i. a device reorientation algorithm, which leverages gyroscope, accelerometer and GPS information, to correct the raw accelerometer data, and ii. a machine-learning framework based on rough set theory to identify rules and detect critical patterns solely based on the corrected accelerometer data. To evaluate the proposed framework, a series of driving experiments are conducted in both controlled and “free-driving” conditions. In all experiments, the smartphone can be freely positioned inside the subject vehicle. Findings indicate that the smartphone-based algorithms may accurately detect four distinct patterns (braking, acceleration, left cornering and right cornering) with an average accuracy comparable to other popular detection approaches based on data collected using a fixed position device.