This paper proposes a force and position hybrid control method for robot manipulators, which makes some of the workspace variables controlled by PD and the other by Kinetic Potential Energy Shaping (KPES). The conventional energy shaping method allows one to design a stabilizing controller with a Lyapunov function candidate consisting of an artificial potential function that plays the role of a design parameter. The potential function depends only on position, and this framework is a natural generalization of the well-known PD control for linear systems. KPES generalizes the conventional method such that it allows one to select an artificial potential function depending on both position and velocity variables. One of the benefits of the conventional method is to preserve the passivity of the original plant system, which enhances the safety of the control system. However, KPES does not preserve passivity in general. This paper proposes a hybrid control method for robot manipulators, which makes some of the configuration variables controlled by PD and the other by KPES. The realized system enables position control of subsystems by KPES and preserves the passivity of the entire system when an external force acts on subsystems controlled by PD. This hybrid control system will be useful from a safety perspective for control problems in which the feedback system still has physical interaction with its environment. This paper presents a method to design a feedback system that guarantees asymptotic stability of the entire system, preserves passivity, and discusses application to a force and position hybrid control with respect to workspace variables of a robot manipulator. Furthermore, numerical examples are presented for a robot manipulator.
Aiming at the challenges that traditional dynamic measurement methods are insensitive to complex trajectory anomalies and conventional filter fitting easily loses motion details, this study proposes a high-precision dynamic measurement method that combines trajectory jump analysis and coordinate sequence processing. This method uses the isolation forest algorithm to automatically identify trajectory jump abnormal points and quantify trajectory deviations. The Transformer model is used to capture long-term motion trends. Finally, the two types of information are fused to generate high-precision measurement results. When the data volume reached 5,000 points, the jump analysis accuracy was 0.97. When the sample batch size was 16, the training time reached 122 s. In actual application, when the global timestamp was 100 s, the highest trajectory error reached 0.14m. At 30 dB, the error increase reached 5.2%. When the rise time reached 0.5 s, the optimal overshoot was 5.1%. The research has higher jump analysis accuracy and calculation speed, as well as better robustness and dynamic response capabilities, providing reliable technical support for industrial automation, intelligent manufacturing and other fields.
Modelling of soil moisture dynamics has attracted attention as a potential solution to lighten the burden imposed on farmers. An accurate soil moisture prediction by the model is necessary to produce an automatic water supply control system, which makes it easy to keep the best soil water conditions for crops. In the case of container cultivation, perlite, which has both suitable water drainage and retention characteristics, is used as soil. However, the water dynamics and distribution in perlite have still unclear. Although there are several models for diffusion dynamics, most of them focus mainly on the direction of gravity. In this work, we study the water dynamics and horizontal distribution of perlite in a container by modelling from experimental data. The experimental results show that the water distribution in perlite over time cannot be expressed as normal diffusion. We propose an extended model that considers the horizontal distribution of the water flux effect and the moisture retention effect. We also discuss the validity of the extended model in terms of water retention and drainage characteristics of perlite.
Decarbonization, decentralization, and digitalization of power systems have been rapidly promoted, but these trends also pose challenges in maintaining supply-demand balance and frequency stability under false data injection attacks (FDIAs). This paper proposes a resilient load frequency control (LFC) system for multi-area power systems capable of maintaining stable power system operation under FDIAs. Specifically, the controller is designed based on a distributed consensus framework with auxiliary variables to suppress system frequency deviations in the presence of FDIAs. It is shown that the proposed system achieves supply-demand balance through consensus on power generation commands, and guarantees Lyapunov stability of the closed-loop system even in the presence of FDIAs. Additionally, our scheme is capable of mitigating any uniformly bounded FDIAs without limitations on the number or locations of compromised controllers, and without relying on attack detection or estimation mechanisms. Finally, the effectiveness of the proposed approach is verified through numerical simulations using the IEEJ 30-machine system model.
This paper proposes an integrated algorithm for matching and pricing in multi-passenger ride-sharing systems. We construct a framework to solve the group formation problem, ensuring that social welfare is maximized even when passengers act strategically. To achieve this, we develop a matching method using a hypergraph structure that allows for the simultaneous optimization of passenger-driver assignments. Additionally, we design a pricing mechanism based on the Vickrey-Clarke-Groves (VCG) theory to guarantee strategy-proofness and individual rationality. Finally, the proposed algorithm is validated through theoretical analysis and numerical simulations.
In this paper, we focus on the weight quantization problem of neural networks, which is essential for practical implementation in control systems. In our previous work, we proposed a weight coefficient quantization method that preserves the intrinsic performance of the neural network as much as possible. However, when a neural network is embedded into a control system, quantization should not only focus on the input-output accuracy but also take into account the system’s dynamic characteristics. To this end, we propose a novel quantization method based on the Markov parameters of the controlled object. Furthermore, we confirm that the proposed approach effectively reduces the network weights while maintaining favorable control performance of the feedforward system through numerical experiments.
In recent years, with the increasing informatization of society, the concept of a super-smart society has been proposed, and technological developments towards its realization are rapidly advancing. The construction of such a society requires a vast amount of information, and technologies such as Wireless Sensor Networks (WSNs) have gained significant attention as effective means of information collection. At this stage, the location information of each node is indispensable for the operation of a WSN; therefore, an appropriate localization system must be designed. The requirements of a localization system include usability in both indoor and outdoor environments, low cost, and low power consumption. Therefore, this aimed to achieve high localization accuracy over a wide area by applying the Apollonius circle locus theory to the received signal strength indicator (RSSI) values from three or more localization reference nodes and by performing maximum likelihood estimation based on a parametric likelihood function. The simulation outcomes demonstrated that the localization error was reduced by approximately 74.8% in a 6m & times;6m area compared to conventional methods, and the 90th-percentile error was suppressed below 0.21m in representative results.
In recent years, many control problems of autonomous mobile robots have been developed. In particular, the robots are required to be safe; that is, they need to be controlled to avoid colliding with people or objects while traveling. In addition, since safety should be ensured even under irregular disturbances, the control for safety is required to be effective for stochastic systems. In this study, we design an almost sure safety-critical control law, which ensures safety with probability one, for a two-wheeled vehicle based on the stochastic control barrier function approach. In the procedure, we also consider a system model using the relative distance measured by a 2D LiDAR. The validity of the proposed control scheme is confirmed by experiments of a collision avoidance problem for a two-wheeled vehicle under vibration.
This paper studies an online automatic parking control problem based on a bidirectional model predictive control (MPC) formulation that considers forward-time motion from an initial state and backward-time motion from a target state simultaneously. By incorporating the terminal condition into the MPC cost function as a soft constraint, the bidirectional formulation becomes suitable for online implementation. However, this relaxation means that the forwardtime and backward-time trajectories are not forced to coincide exactly. To compensate for this mismatch, a complete online control scheme is proposed based on the bidirectional MPC with online reversal-point determination and trajectory tracking. Numerical examples show that the proposed method can generate feasible parking maneuvers while determining the reversal point online, even when the forward-time and backward-time trajectories do not merge exactly. The results also indicate that appropriate weighting design in the MPC cost function is important for shaping vehicle behavior and naturally reproducing parking motions consistent with human driving behavior.
Random packet losses and transmission delays during data transfer for bilateral control systems can cause serious problems such that the slave device cannot perform the desired operation, and the operator cannot take the status of the slave side via the master device. This paper is focused on bilateral control using flexible structures as slave devices, and presented a method of not only disturbance removal, but also the estimation and data compensation that takes into account random packet losses and transmission delays by using a Kalman filter algorithm. Moreover, numerical simulations show that the proposed method can handle continuous operation with a 50% packet loss rate.
This paper addresses the design of an assistance control system that enables safe and flexible operation while facilitating operator upskilling. A key factor in upskilling is ensuring operators have repeated opportunities to engage in control actions. To this end, we design an assistance system that reduces intervention frequency and promotes operator discretion. We particularly focus on designing a human assistance system for water-level control in a tank system. Then, we characterize the set of admissible control actions for the human operator to develop assistance control logic that ensures the safety of the overall system. Finally, we demonstrate water-level control using the proposed assistance control logic. The results verify that the proposed system ensures the safety of the overall system with reduced intervention, highlighting its potential as a foundational technology for effective operator upskilling.
Soil compaction is a critical element in construction, as it directly influences the quality and durability of structures. Vibratory rollers are widely employed to enhance ground stiffness, with traditional methods emphasizing the number of compaction cycles. However, these methods require preliminary testing to establish the relationship between compaction cycles and ground stiffness, failing to account for actual ground conditions that vary by location. Continuous compaction control (CCC), utilizing accelerometers mounted on vibratory rollers, has been introduced as a more effective quantitative evaluation method tailored to real-world ground conditions. This approach measures the distortion rate of the acceleration waveform generated by the vibratory roller. Nevertheless, since the accelerometer is affixed directly to the drum, the measurement is inevitably affected by noise from the vibration source. To address this limitation, this study proposed an innovative ground stiffness evaluation method employing multiple accelerometers installed on the ground surface. This proposed technique is significantly less influenced by noise directly caused by the vibratory roller. Experimental results demonstrated that the proposed method offers superior suitability for quantitatively assessing ground stiffness. By adopting this method, a more precise evaluation of ground stiffness can streamline quality control processes and minimize the need for rework.
This study investigates a periodic-review inventory system for perishable products with a fixed shelf life. Demand in each period is stochastic and nonstationary. Given the complexity of determining an optimal inventory control policy, prior research has introduced the nested marginal cost accounting scheme, which enables effective evaluation of order quantities based on their marginal costs. Building on this scheme, we develop an approach that approximates an optimal policy by minimizing the sum of expected marginal costs associated with the order quantity for the current inventory state. To this end, we derive the partial derivative of this sum with respect to the order quantity, whose evaluation is less computationally demanding than the original cost function. This partial derivative is incorporated into Brent’s method to achieve efficient minimization. We further extend our approach to systems with positive setup costs by integrating a well-known heuristic that jointly determines both the ordering cycle and the corresponding order quantity. Our computational experiments compare the proposed algorithms with existing algorithms. The results demonstrate that our algorithms provide superior solution quality and computational efficiency.
Food cutting is an essential skill of housekeeping robots. The cutting motion requires the stable holding of food items. In this study, we propose a search method for finding a stable holding point set during the cutting process. The fracture and friction forces generated by the knife were computationally estimate at a certain pose. Additionally, we discuss the calculation of the holding force corresponding to the holding point set to counteract the knife wrench. Then, an evaluation function is used to score potential holding point sets based on the magnitude and direction of the holding force. Finally, we search for the holding point set with the highest score as the final output. The proposed method was evaluated on our dual-arm robot system equipped with a two-finger hand. The results indicate that the identified holding point sets enable stable grasping during the cutting situation.
This study aims to conduct effective analysis on the physical training data of civil aviation student pilots, thereby accurately analyzing the changing trends of students' physical fitness status and providing a scientific basis for optimizing training programs. First, the Iterative Dichotomiser 3 (ID3) decision tree algorithm is used to select key features from flight students' physical training data by calculating information gain and classify the data. Then, the Back Propagation Neural Network (BPNN) algorithm with strong nonlinear fitting capability is employed to further analyze and predict the classified data. This fusion method aims to combine the interpretability of decision trees with the pattern recognition ability of neural networks. The dataset from the Physical Training Center of Civil Aviation University of China is used to verify the algorithm performance. The results show that the classification accuracy of the ID3+BP algorithm is 92%, higher than other algorithms. The algorithm can accurately identify the impact of key features such as muscle strength and endurance level on physical fitness status, making the classification results closer to the actual physical fitness status and providing a reliable basis for personalized training programs. In the analysis of physical training data, the ID3+BP algorithm achieves a precision of 90%, a recall of 89%, and an F1 score of 89.5%, all outperforming other algorithms. In conclusion, this study provides a new method for analyzing physical training data of flight students, showing great potential for improving their physical fitness and flight safety.
This paper addresses the design of an energy management system (EMS) that embeds a mechanism to encourage cooperative behavior among electric vehicle (EV) users. The proposed EMS provides incentives for EV users to contribute to grid stability by staying at home and allowing their vehicles to discharge electricity, thereby serving as a reserve power source. To this end, we construct a model that captures the behavioral response of EV users to the offered incentives. Then, we formulate the problem of incentive-driven EMS (ID-EMS) design as an optimization framework based on the constructed user behavior models, and present a policy for determining the optimal incentive in practical operation. Finally, we evaluate the performance of the ID-EMS through numerical simulations, utilizing real-world datasets on daily vehicle usage and behavioral insights obtained from user surveys.
In this paper, we propose a method for constructing parametric control Lyapunov functions (parametric CLFs), which depends on a parameter & varepsilon;. The parametric CLF is constructed by using a notion of symmetry allowing state-space transformations, time-scale transformations, and feedback transformations. In this approach, a nominal CLF is deformed by these transformations, and the deformed functions are also CLFs of the original system. The resulting parametric CLF varies in shape depending on the parameter & varepsilon;\epsilon. A parametric CLF with a small & varepsilon; is suitable for fast behavior, while a CLF with a large & varepsilon; is suitable for slow behavior. We let the parameter & varepsilon; depend on the value of the CLF, which yields an implicit equation. Under a uniqueness condition of the implicit equation, the resulting implicit function qualifies as a CLF. When the state is near the origin, the implicit CLF is suitable for fast behavior; when the state is far from the origin, the CLF is suitable for slow behavior. Furthermore, we show that, for a class of systems, a modified Sontag's universal controller using the implicit CLF ensures bounded input signals.
This paper proposes an adaptive detection method for false data injection attacks based on the retraining of recurrent neural networks (RNNs). Machine learning-based detection methods heavily depend on historical pretraining data. Because of this dependency, the detection performance degrades when the power system largely fluctuates and becomes unexpected states. We propose retraining with new data obtained under unexpected states to avoid this serious degradation. To fasten the retraining, well-known RNNs such as long short-term memory (LSTM) and gated recurrent unit (GRU) should reduce the number of epochs, which prevents them from achieving high performance. To overcome this tradeoff, we utilize the echo state network (ESN), which is a representative model of Reservoir Computing. ESN significantly reduces training time compared with LSTM and GRU. We evaluate the proposed method with the three neural networks using the IEEE 68 bus system, confirming the performance of detection significantly degrades when the system enters an unexpected state. We also show the proposed retraining is effective in recovering the performance, and ESN outperforms the other RNNs in terms of retraining time.
In the emerging aging society, the expectation for sufficiently safe and reliable mobility tools, like automated wheelchairs, is growing to keep the lives of physically needed persons active and comfortable. An essential intelligence required for an automated wheelchair is interacting with pedestrians in a crowded pathway. This paper addresses this issue and proposes an optimal control approach for interactive trajectory generation based on a pedestrian behavior model in typical pedestrian interaction scenarios. Specifically, we propose a novel way to incorporate pedestrian behavior near the wheelchair using the social force model (SFM) under the predictive optimal control framework without requiring extra computational burden. Such incorporation of SFM enables the wheelchair to intelligently predict the multiple pedestrian interactions when the wheelchair emerges in their range of vision in complex environments. Finally, control decisions in the linear model predictive control framework, which considers conditional safety constraints, ensure optimal performance even in adaptive or changing pedestrian environments.
