不精準資訊下基於到達時間差之合作式車載通訊定位技術

Abstract

在 V2V 場景下的合作式定位是一個被廣泛討論的技術，若是各個系統之間能彼此提供可靠的資訊，通過與鄰居共享位置資訊，定位的精準度能提升不少。在本論文中，我們設計了一個結合卡爾曼濾波器與運動，GPS 和到達時間（TOA）的混合模型。在這樣的混合模型之下，我們研究如何增加資訊進入卡爾曼濾波器更新過程以改善有限的定位性能。我們透過 TOA 的原始數據彼此相減生成新的一組到達時間差（TDOA）的測量值，嘗試增加整體系統的資訊以改善定位結果。雖然生成的 TDOA 來自原本就存在的TOA 測量值，但TDOA 測量值通過在卡爾曼濾波器更新過程中使用不同的 Jacobian 矩陣，就像以一種不同視角對系統提供額外的資訊。在卡爾曼濾波器的更新之中，測量值模型的方差會透過影響更新步伐大小來影響整體收斂的結果，在這樣的前提之下，我們透過考慮鄰居的不確定位置資訊來得到一個新的TDOA 模型方差; 而考慮了鄰居不確定位置方差，能為卡爾曼濾波器更新的每個測量值帶來適當的更新步伐大小。最後我們把原始的 TOA 測量值還有生成的 TDOA 測量值一起放進卡爾曼濾波器中更新，模擬的結果顯示，混和兩種測量值的更新效果會比只使用原始TOA 的平均定位效果好 15%。

Cooperative localization for V2V communications are widely be discussed and used in many situations. By sharing position information with neighbors, we could improve the localization performance under the premise of the credible data. In the thesis, we design an localization algorithm using the Kalman lter with the fusion model of motion, GPS and the time of arrival (TOA) measurement. Under such fusion model, we wonder if there is any possible information can be added into the Kalman update process to improve the limited performance. We construct the generated time different of arrival (TDOA) measurement set from the original ranging measurement try to put more information to improve the localization performance. By using different Jacobian matrix in Kalman update process, the generated model provides a different perspective of the original measurement data. However, under the V2V scenario, the ranging measurement TOA is suffered from the uncertain neighbor position so as the generated measurement from TOA. It is well known that the variance of the measurement model have a significant impact on the Kalman filter performance in estimating dynamic states. We modify the variance of the generated measurement model by considering the uncertainty of the neighbor position to give the appropriate update step size for the Kalman filter. After the neighbor uncertainty consideration of the Kalman update process for generated measurement, we put the generated measurement into the non-linear update process together with the original TOA measurement set to form a hybrid ranging measurement update system. The results show that the fusion model with hybrid measurement set can improve the localization performance around 15 % of the average error with the suitable GPS level. The proposed generated measurement with the modified variance can improve the localization performance without other additional communication devices.