基于时序伪距差分的融合定位方法研究

Navigation and location services has become an indispensable modern technology in our social life. However, the traditional technology is still helpless in many situations such as the city and the indoor environment. A new method based on the timing pseudorange differential positioning approach is proposed by converging the principle of dead reckoning, TDOA (Time Difference of Arrival) positioning model and GNSS (Global Navigation Satellite System) differential approach. The step of the adjacent epoch is estimated by the time difference of the pseudorange, and the change of the position and the trajectory of the carrier are estimated by the step and the heading angle. This method can effectively eliminate the common error of the ranging signal and the cumulative error generated by the estimation to ensure the accuracy of continuous positioning. The method can realize continuous navigation based on single anchor point, and when the number of anchor points increases, it can use redundant information to improve the positioning accuracy and reliability effectively. Combining the timing pseudorange difference method with Kalman filtering technique, the positioning results can be further improved. The principle of pseudo-distance difference positioning method is clear and flexible, and it breaks through the constraints of GNSS four-star intersection positioning principle, reduce the accumulate error of the dead reckoning methods and the accuracy requirements of time synchronization. Projection, TOA (Time Of Arrival) positioning and differential principle, has a good universal application value.

导航定位和位置服务已成为社会生活中必不可缺的现代化技术。然而在城市及室内等诸多场合，传统技术仍然存在大量无法应用的盲区。通过融合航位推算原理、到达时间差(TDOA)定位模型及全球导航定位系统（GNSS）差分思想，提出了一种时序伪距差分定位方法。通过时序伪距差分推算相邻历元用户步距，结合其他定位信号（如行进方位角）不断推算载体位置坐标变化与行进轨迹。本方法有效消除测距信号的公共误差以及推算累计误差，保证连续定位的精度。基于单锚点即可实现连续导航，而当锚点数量增加时，能够利用冗余信息有效提高定位精度与可靠性。将时序伪距差分方法与卡尔曼滤波技术结合，对误差进行建模和处理，可以进一步优化定位结果。时序伪距差分定位方法原理清晰、使用灵活，突破了GNSS四星交会定位原理约束，弥补了航位推算累计误差影响导航精度的缺陷，降低了对锚点时间同步的精度要求，具有良好的普适应用价值。

室内外复杂导航环境下的高精度鲁棒定位已成为当前诸多前沿领域的迫切需要。大部分应用场景均依靠采取高精度传感器融合方法，以保证定位精度及鲁棒性，但是其高昂的成本限制了此类方案的规模化应用。定位导航采用的传感器几乎都属于时间序列观测，这些序列观测数据提供了系统模型观测的诸多约束信息，本项目针对时间序列观测数据的特点，通过采用在线深度学习网络对相关参数进行建模，并结合不变观测量理论和自适应滤波理论，建立了基于时序伪距差分模型的自适应滤波框架，实现灵活的传感器数据融合，在室内外复杂导航环境下，有效的保证了滤波精度和鲁棒性。研究结果表明，基于序列数据驱动的建模方法，相比于传统的基于理论假设的模型，具有更好的环境适应性和抗差能力，通过在良好环境下传感器所采集的数据对模型进行反复的训练和修正，并在复杂导航环境下通过共享这些参数，实现以较低成本获得鲁棒的高精度定位能力。本项目的滤波模型在行人航位推算、室内多传感器融合定位等实际应用中均取得了较为显著的提升，证明了该研究思路的可行性及研究和应用价值。