Abstracts Track 2023

Area 1 - Industrial Informatics

Nr: 192

A Robust Path Planning Approach by Vision-Based State Observation for FDM Type 3D Printing Process


Shinichi Ishikawa, Takahito Yamashita and Ryosuke Tasaki

Abstract: In the deposition-based 3D printing process, the material stacking error during printing is a major defect that affects the entire process, therefore a robust printing method that compensate the nozzle movement path during the printing process is needed. The main purpose of our research is to demonstrate the effect of dynamic compensation control for path changes during the printing process of column shape by observing the around the nozzle using cameras and compensating the nozzle path using vision-based real-time feedback control. In this approach, two cameras are used to avoid the problem of occlusion. The whole area could be monitored, and ROI (Region of Interest) is defined to observe in a specific area. The RBG images from two cameras are cropped and combined to the desired size. Additionally, binarization and edge detection are applied to the ROI. An AR marker is placed at the center of the print target to obtain the center coordinate. The feedback controller acquires the distance between the center coordinates and the target path in real-time and calculates a compensation motion according to the error from the target path. In this research, a printing experiment have been conducted using a material for verification, assuming layering. In the case of 3D printing experiment for a column shaped object with a radius of 50 mm, and the compensation effect has been verified when a convex path change occurred in a target path. The mean absolute percentage error (MAPE) between the motion path and the target path is 3.62 %, and the result indicates the effectiveness of the failure suppression method by correcting the path during the printing process. On the other hand, the RMSE and maximum error value of the experimentally obtained paths are 2.80 mm and 8.94 mm respectively, the large error is caused by detecting unexpected area in the image processing process. As a future works, a new path modification method to target the depth images only in specific layers is implemented. The main advantage is that the high dimensional printing accuracy of final product is not to relate to the printing errors in the lower layers or the errors in the background.

Nr: 195

A State-Space Approach to Study the Bullwhip Effect in Supply Chains Under the Rationing Game


Christos Papanagnou

Abstract: Motivated by the disruption phenomena in supply chains when demand for goods exceeds supply, a stochastic state space model is introduced to study the impact of the rationing game on the bullwhip effect and inventory variations. The rationing game is modelled with the aid of a proportional gain factor, which adjusts the number of goods dispatched to the downstream supply chain nodes. A two-echelon supply chain is considered in this study, which consists of a single distributor that serves two retailers. It is assumed that inventory replenishment in each retailer follows base stock policies under customers' stochastic demand profiles. The dynamic properties of the supply chain model are encapsulated in a closed form covariance matrix, which is expressed as a function of the proportional control parameters and the percentage of the distributor's inventory that is dispatched to the retailers. The model is analysed under stationary conditions, allowing to analyse the effect of distributor's inventory variances and correlated demand profiles on the bullwhip effect (demand amplification) and related instability phenomena in supply chains.

Area 2 - Intelligent Control Systems and Optimization

Nr: 191

Robust Extremum Seeking Control and Optimization for Vehicle Platooning


Yasser Bin Salamah

Abstract: In this work, we investigate the application of extremum-seeking control (ESC) techniques in the context of vehicle platooning to optimize the efficiency and safety of platoon operations. Vehicle platooning, where multiple vehicles travel closely together, offers potential benefits such as improved traffic flow, reduced fuel consumption, and increased safety. The primary objective of this research is to develop a robust sliding mode-based ESC framework specifically designed for vehicle platooning. The framework will enable platoon vehicles to autonomously seek and track optimal operating points, such as optimal inter-vehicle spacing and speed, to achieve improved fuel efficiency while maintaining safety constraints. Inspired by our previous work in cooperative extremum seeking, the proposed framework will employ a combination of mathematical modeling, optimization algorithms, and real-time control strategies in a cooperative fashion. By incorporating feedback from sensors, vehicle-to-vehicle communication, and platoon coordination algorithms, the ESC framework will adaptively adjust the platoon formation to optimize fuel efficiency and safety in dynamic traffic scenarios. The work will illustrate the applicability of the proposed control through extensive simulation studies using advanced simulation tools. Key performance metrics, including fuel consumption, travel time, and safety measures, will be evaluated to assess the effectiveness of the ESC approach compared to conventional platooning strategies. The anticipated outcomes of this research include enhanced fuel efficiency, reduced emissions, and improved safety in vehicle platooning operations. The findings will contribute to the development of more efficient and sustainable intelligent transportation systems, particularly in the context of connected and autonomous vehicles. By leveraging extremum-seeking control techniques in vehicle platooning, this research aims to advance the understanding and implementation of optimized platoon operations, promoting more efficient and environmentally friendly transportation solutions.

Nr: 193

Reinforcement Learning with High Efficiency and Low Computational Effort in Actor-Critic Problems over Infinite Horizon


Vincenzo Basco

Abstract: Reinforcement Learning (RL) is a branch of machine learning that deals with training agents to make decisions in complex environments. It has applications in various domains, including robotics, gaming, and autonomous systems. RL involves an agent interacting with an environment, taking actions to maximize a cumulative reward signal over time. Actor-Critic methods are introduced as a promising approach within RL in solving Hamilton-Jacobi-Bellman equations. The two fundamental components: the "Actor" and the "Critic". The Actor is responsible for learning a policy, which defines the agent's behavior by specifying which actions to take in different states. On the other hand, the Critic aims to estimate the value of being in a particular state or taking a specific action. This combination of policy-based (Actor) and value-based (Critic) learning helps address challenges posed by the lack of regularity of weak solutions, whenever state constrains are imposed ([1]). The notable aspect highlighted in the speech is the use of weak solutions regularity ([2]) and Hartley algebras to tackle Actor-Critic problems in RL. Hartley algebras are mathematical structures that have properties useful for solving RL challenges. Specifically, we show that Hartley algebras can aid in designing advanced exploration strategies for approximating the value function in RL with Low Complexity effort. These strategies can improve the efficiency of the learning process and the agent's ability to adapt to complex environments. More specifically, we introduce a novel tuning law for critic weights in value approximation. This is based on solving an equation involving various parameters, including the state, control input, critic weight vector, neural network activation vector, and a bias term. Notably, this tuning law deviates from the commonly used Levenberg-Marquardt algorithm, which is often employed to solve nonlinear least squares problems. References: [1] V.B. Lipschitz regularity of controls and inversion mapping for a class of smooth extremization problems, Automatica. Vol. 148, February, 2023. [2] V.B. Weak Epigraphical Solutions to Hamilton-Jacobi-Bellman Equations on Infinite Horizon, Journal of Mathematical Analysis and Applications, Volume 515, Issue 2, 15 November 2022 (126452), 2022.

Area 3 - Robotics and Automation

Nr: 194

Multi-Sensors Based Intelligent Situational Awareness and Cooperative Control System of Autonomous Ships in a Marine Environment: Development and Preliminary Field Tests


Jeonghong Park, Jinwoo Choi, Minju Kang, Kibeom Choo, Namhoon Ha and Hyun-Taek Choi

Abstract: This paper introduces the development of intelligent situational awareness and autonomous navigation systems, which are core systems required for a maritime autonomous surface ships (MASS) and an autonomous surface vehicles (ASV), representative unmanned platforms that maneuver autonomously at sea, and the preliminary field test results of the developed systems. First, one of the core systems is a multi-sensor fusion-based situational awareness system that automatically detects and identifies static and dynamic objects located around the ships (MASSs and ASVs) during autonomous navigation and generates predictive information on dangerous situations such as collisions. The developed system considers the characteristics of perception sensors such as cameras, lidar, and radar, and includes the capabilities to automatically detect maritime objects through linkage with navigation equipment. In order to maintain stable detection performance despite rapidly changes and variations in the maritime environment, useful training data were obtained and artificial intelligence techniques were implemented. An extended Kalman filter-based sensor fusion technique was then applied to estimate the position, speed, and course information of the detected objects, which was used to evaluate a quantitative indicator of the potential collision risk of the ship’s predictive course. The applicability of the developed system and the proposed techniques were validated using data obtained from real-world maritime environments. Second, another system for improving autonomy is an autonomous navigation system that considers cooperative navigation with MASS and/or ASVs. To demonstrate cooperative autonomous navigation algorithms and systems, we developed three catamaran-shaped ASVs, including guidance, navigation, and control (GNC) systems, electric propulsion, and communication network systems. In particular, for cooperative navigation between multiple ASVs, a network communication system was implemented to synchronize and share motion information for each vehicles through a multi-master system in robot operating system (ROS). For autonomous navigation, we implemented a cooperative navigation and control scheme based on the relative geometry information between ASVs in the developed autonomous navigation framework to follow a predefined path and maintain a specific formation based on the relative distance between ASVs. Herein, the formation control scheme was designed based on the artificial potential field approach with virtual leader and followers structure. Moreover, to demonstrate the fundamental maneuvering performance and the practical feasibility of the developed cooperative navigation and control scheme with the three ASVs, preliminary field tests were conducted in an inland water environment, and the test results introduce in the study.

Area 4 - Signal Processing, Sensors, Systems Modelling and Control

Nr: 19

The Stability of a Linear Time-Invariant System With Control Delay: Application to the Stability of the Aircraft Control Chain in Conditions of Atmospheric Turbulence


Daniela Enciu, Adrian Toader and Ioan Ursu

Abstract: A class of time-invariant linear systems with control delay and additive disturbances is considered. Through a predictive state feedback method, the control delay is compensated, reaching a closed-loop system without delay. Based on a theorem of F. Mazenc, S.-I. Niculescu, M. Krstic, stability is ensured in the presence of disturbances. The application is made on the control chain of an airplane, in the presence of Dryden-type atmospheric turbulence.

Nr: 196

Online Estimations for Supercapacitors Using Hybrid Nonlinear Observers


Eric Magarotto, Tarek Ahmed-Ali and Madjid Haddad

Abstract: SuperCapacitor (SC) is one solution as Energy Storage System with batteries, fuel cells and flywheels, each of them have specifics properties depending on their use. In Electric Vehicle (EV), they can be used along the main battery to benefit from their complementary characteristics. As technology continues to advance, exploiting their high power density, wide operating temperature range and long cycle life, their use in various industries and applications is likely to grow. Recently, Supercapacitors have undergone various improvements which make it possible to extend their use, thus causing a resurgence of interest in the study of SCs. In the case of EV, the SC is part of a Management Systems (MS) to optimize energy usage, extend supercapacitor lifespan, and enhance overall system performance. For the MS, it is necessary to provide SOx (x=charge, health, energy) informations based only on a few measurements (voltage, current). To prevent overheating, voltage pikes or abnormalities, the MS need real-time monitoring, then an estimation of useful parameters. Each online method has advantages and disadvantages depending on the objectives. Among these, we can find observer-based, filtering-based, data-driven techniques. The observer-based methods achieve high accuracy and have good computational efficiency, which makes them suitable for embedded implementation in MSs. Notable results are obtained by the use of an UKF, GESO or hybrid observer. To avoid a calculation overload, they are obtained either with very simple models and fairly simplistic current profiles. For model-based techniques, quality and complexity of the model are crucial points. From a computational efficiency and implementation point of view, an Equivalent Circuit Model model is suitable. The simplest model failed to reflect some important phenomena (self-discharge, diffusion behaviour) but is enough in an ageing goal with stepwise current profile (Magarotto et al, “A New Observer Design for Aging Detection of Supercapacitors”, JDSMC 2019). It turned out to be a better alternative than the Spectroscopy method. To better reflect EV reality, new current profile (from NEDC or WLTP norms) must be adopted. This is quite rare in observer-based research work, because of the strong dynamics of the input and the resulting difficulties, even more considering a larger sampling time. The use of an inter-sampled observer-predictor, with the aid of Lyapunov theory, has shown, in a different context (Magarotto et al, “A new sampled-data observer design for bioreactors”, CODIT22), that it could provide solutions to this kind of problem. Applying it to Supercapacitors, recent results were obtained with real data but with simple model (Magarotto et al, “Sampled-data observer for supercapacitor parameters estimation under NEDC cycles”, CODIT23). Our current work focuses on an extension to multi-branch models. First results give good estimates of voltages with a gain tuning for the trade-off between sampling–time and performances. Our actual objective is to design an hybrid exponentially convergent observer to achieve the same quality of parameters estimation with NEDC/WLTP profiles while respecting the minimum core consumption (increasing sampling period). Finally, they are still many challenges to overcome: one must explore robustness (to measurement noise) and the use of a diffusion model which lead to a natural extension of observers for PDE.