研究目的
To estimate distance using a sensory fusion model combining three reflective optical distance sensors of different ranges, a color sensor (VIS), an ultraviolet radiation sensor (UV), and a Near infrared sensor (NIR) with a Multi Layer Perceptron (MLP), compensating for radiation exposure to adjust the measurement.
研究成果
The implementation of sensor fusion models based on MLP significantly improves the precision of distance measurements, achieving a mean square error lower than 1%. The study demonstrates the MLP's capability to compensate for external radiation disturbances, with the Tangential Sigmoid activation function providing better performance than the Linear function. The method allows for implementation in low-cost embedded systems, enhancing data sampling frequency and processing efficiency.
研究不足
The study acknowledges the challenges of sensor fusion, including uncertainty and imprecision due to environmental conditions, device failures, communication issues, and climatic conditions. The performance of low-cost reflective optical sensors is affected by nonlinear response, limited operating ranges, and sensitivity to infrared or visible radiation.
1：Experimental Design and Method Selection:
The study employs a Multi Layer Perceptron (MLP) for sensory fusion, combining data from three reflective optical distance sensors, a color sensor (VIS), an ultraviolet radiation sensor (UV), and a Near infrared sensor (NIR).
2：Sample Selection and Data Sources:
Data acquisition was performed using a linear actuator as a reference, with experiments conducted under different external radiation sources affecting the distance sensors' measurements.
3：List of Experimental Equipment and Materials:
Includes reflective optical distance sensors (GP2Y0A41SK0F, GP2Y0A21YK0F, GP2Y0A02YK0F), a color sensor (VIS), an ultraviolet radiation sensor (UV), a Near infrared sensor (NIR), and a high-precision linear displacement system.
4：Experimental Procedures and Operational Workflow:
The process involved taking 20 samples every 1 cm, moving the system in the range between 3 cm to 90 cm, under four different light sources: Halogen 50 W, Incandescent 40 W, Incandescent 50 W, and without light presence.
5：Data Analysis Methods:
The MLP was trained with different architectures (varying the number of hidden layers and neurons per layer) and activation functions (Tangential Sigmoid and Linear) to evaluate performance based on Mean Square Error (MSE).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
