A Novel Friction Compensation Method Based on a Hybrid Model for Vision-Based Optoelectronic Platform

摘要

With the development of technology, the demand for detection distance and stable tracking accuracy of vision-based stable tracking platforms is getting much higher. However, the line of sight often shakes due to friction, which ultimately causes a decline in imaging quality and a reduction in stable tracking accuracy. Inspired by Lurge friction model and neural network, a hybrid friction model has been proposed in the paper to improve the performance of the photoelectric platform. First of all, a Lurge friction model is established for the static friction part and its parameter is identified by means of differential evolution algorithm. Second, a recurrent single multiplication neuron is built for the system's dynamic friction part and the differential evolution algorithm is also introduced to obtain the model's parameters based on the speed change data collected during free deceleration movement. Finally, the experimental results show that the proposed hybrid friction model can effectively approximate the realistic friction and eliminate the "flat top" and "dead zone" phenomena of speed response. The maximum speed tracking accuracy is reduced to ±0.05 °/s after compensation, which greatly improves the system's speed tracking accuracy and stability accuracy.