Batteries play a critical role in electric vehicle systems devices. The safety and performance of these applications rely on accurate Battery Management Systems (BMS) to monitor and optimize battery performance. Traditional BMS systems face challenges in charge prediction processes due to complex chemical processes and battery aging, leading to faults. The absence of a perfect sensor highlights limitations in measurement issues arising from external factors, especially sensor noise. Therefore, there is a need for algorithms that can solve the real-world battery charge prediction problem. This study compares an innovative solution, the Transformer model, with traditional Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM) models, and Support Vector Regression (SVR). This research aims to provide new perspectives on Battery State of Charge (SoC) predictions using NASA, BMW i3, Stanford University Battery Datasets, and real-world battery data obtained from L5 electric vehicles of the Musoshi brand collected for this work. This research’s primary objective is to apply the Transformer model to real-world battery data, evaluating it as a significant step in electric vehicle range optimization and battery management. The Transformer model in the study achieved the best result with an RMSE value closest to 1 (~0.99).
Batteries are essential for Electric Vehicles (EVs). Traditional Battery Management System (BMS) algorithms can be inadequate for State of Charge (SoC) estimation due to incorrect measurements and unobservable battery characteristics. Centralized machine learning methods are used to improve SoC estimation. Both privacy and high bandwidth requirements are the main disadvantages during the implementation. Federated Learning (FL) solves these issues by performing local learning on devices, protecting data privacy, and only aggregating model updates at the central server. While FL approaches can help preserve data privacy during model training, collaborative learning can facilitate the integration of priori data learned by the agents in the fleet with the rest of the fleet members to improve charge prediction. In this study, a new aggregation rule named Federated Adaptive Client Momentum (FedACM) is proposed to handle data imbalance and heterogeneity in SoC estimation. The proposed method is initially validated via experimental results using the collected data from Musoshi L5 type EV. It is also tested using publicly known datasets such as NASA, BMW i3, and Standford University Battery Datasets. These experiments show that the proposed aggregation rule performs better than current state-of-the-art rules for FL.
Bipedal locomotion has good obstacle avoidance properties. A robot with human appearance has advantages in human-robot communication. However, walking control is difficult due to the complex robot dynamics involved. Stable reference generation is significant in walking control. The Linear Inverted Pendulum Model (LIPM) and the Zero Moment Point (ZMP) criterion are applied in a number of studies for stable walking reference generation of biped robots. This is the main route of reference generation in this paper too. We employ a natural and continuous ZMP reference trajectory for a stable and human-like walk. The ZMP reference trajectories move forward under the sole of the support foot when the robot body is supported by a single leg. Robot center of mass (CoM) trajectory is obtained from predefined ZMP reference trajectories by Fourier series approximation. We reported simulation results with this algorithm in our previous works. This paper presents the first experimental results. Also the use of a ground push phase before foot take-offs reported in our previous works is tested first time together with our ZMP based reference trajectory. The reference generation strategy is tested via walking experiments on the 29 degrees-of-freedom (DOF) human sized full body humanoid robot SURALP (Sabanci University Robotics Research Laboratory Platform). Experiments indicate that the proposed reference trajectory generation technique is successful.