基于BP神经网络的车辆定位融合模型

Fusion Model of Vehicle Positioning Based on BP Neural Network

针对常规的GPS定位方法在大城市中容易丢失信号的问题 ,提出采用GPS与基于移动通信网络的定位技术 (MPS)相结合的思路 ,利用反向传播 (BP)神经网络构造GPS与MPS定位信息的融合模型 ,并采用动量法和学习率自适应调整的策略来解决BP算法收敛速度慢和局部极小点的问题 .用 1 2 6条调查数据对神经网络进行训练的结果表明 ,该模型结果在定位的方向和趋势上基本与GPS定位结果一致 ,且不依赖于原有模型 。

Signals are prone to be lost in big cities when using traditional GPS. To solve this problem, a concept which combines GPS and mobile-communication-ne twork-based MPS was adopted, and the BP neural network was used, thus constructi ng a vehicle positioning fusion model of GPS and MPS positioning information. By utilizing the momentum method as well as the self-adaptive adjusting strategy o f learning rate, the slow convergence speed of BP algorithm and the local minima l point were solved. Training results with 126 items of research data on the net work indicate that the model is consistent with GPS both in direction and in tre nd and is independent upon the traditional GPS model. So the model can efficient ly be applied in maintaining the positioning continuity and precision with lower cost.