Research on Multi-Sensor Data Fusion for Smart Elderly Care Robots

Against the backdrop of increasing demand for elderly care, elderly care robots play a special role. Currently, the multi-sensor data fusion in elderly care robots faces issues such as low accuracy of sensor data collection and the need for optimization and innovation in fusion algorithms. To address these issues, this paper optimizes the fusion algorithm that combines the Kalman filter algorithm, artificial neural network (ANN) algorithm, and SVM algorithm. First, raw data is collected from the sensors of the elderly care robot, checked, cleaned, and reprocessed using min-max normalization. Then, the Kalman filter performs preliminary data processing, state prediction, and calculates relevant parameters to reduce noise and uncertainty, improving accuracy. Next, the artificial neural network receives the filtered data, proportionally divides the data set, selects ReLU and linear activation functions for the hidden layer and output layer respectively, uses mean squared error as the loss function, and updates weights using back propagation and stochastic gradient descent algorithms to input data into the fusion to obtain feature vectors. Finally, SVM uses this vector as input, trains the model with labeled data, inputs the newly collected and processed sensor data into the model for prediction, and adjusts parameters based on the results to achieve final fusion and decision-making. An experimental environment is established in nursing homes or senior living communities, and comparative tests with other fusion algorithms show that this algorithm has greater advantages in the accurate collection of sensor data for elderly care robots.