From 60be87c3f92fd3cd9b3b55cec672381d84ae342a Mon Sep 17 00:00:00 2001 From: ZhouJiafen Date: Tue, 29 Dec 2020 00:39:17 -0800 Subject: [PATCH] =?UTF-8?q?=E6=9B=B4=E6=96=B0imu=E4=B8=ADbias=E8=BF=AD?= =?UTF-8?q?=E4=BB=A3=E7=9A=84=E6=B3=A8=E9=87=8A=E4=BB=A3=E7=A0=81?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/ImuTypes.cc | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/ImuTypes.cc b/src/ImuTypes.cc index 84336ac..d90d609 100644 --- a/src/ImuTypes.cc +++ b/src/ImuTypes.cc @@ -315,7 +315,8 @@ void Preintegrated::IntegrateNewMeasurement(const cv::Point3f &acceleration, con B.rowRange(6,9).colRange(3,6) = 0.5f*dR*dt*dt; // Update position and velocity jacobians wrt bias correction - // ? 更新bias雅克比,计算偏置的雅克比矩阵，pv 分别对ba与bg的偏导数,论文中没推这个值，邱笑晨那边有推导 + // ? 更新bias雅克比,计算偏置的雅克比矩阵，pv 分别对ba与bg的偏导数,论文中没推这个值，邱笑晨那边也没有推导, + // 但论文作者对forster论文公式的基础上做了变形，然后递归更新，参见 https://github.com/UZ-SLAMLab/ORB_SLAM3/issues/212 // 因为随着时间推移，不可能每次都重新计算雅克比矩阵，所以需要做J(k+1) = j(k) + (~)这类事，分解方式与AB矩阵相同 JPa = JPa + JVa*dt -0.5f*dR*dt*dt; JPg = JPg + JVg*dt -0.5f*dR*dt*dt*Wacc*JRg; @@ -341,6 +342,7 @@ void Preintegrated::IntegrateNewMeasurement(const cv::Point3f &acceleration, con // Update rotation jacobian wrt bias correction // 计算偏置的雅克比矩阵，r对bg的导数，∂ΔRij/∂bg = (ΔRjj-1) * ∂ΔRij-1/∂bg - Jr(j-1)*t + // 论文作者对forster论文公式的基础上做了变形，然后递归更新，参见 https://github.com/UZ-SLAMLab/ORB_SLAM3/issues/212 JRg = dRi.deltaR.t()*JRg - dRi.rightJ*dt; // Total integrated time