Abstract: Disclosed are an attitude estimation method, a terminal, a system and a computer-readable storage medium. For the problem of attitude and sensor bias estimation, a cascade solution method is provided. The first part of the cascade is a Kalman filter applied to an LTV system, and the second part of the cascade is a nonlinear attitude observer built in SO(3). In the estimation process, only one constant inertial reference vector needs to be measured explicitly in body-fixed coordinates, and the complexity of the attitude estimation algorithm is greatly simplified by exploiting the geometric relationship between the inertial reference vector and the Earth angular velocity vector. At the same time, the time-varying characteristics of the implemented Kalman filter help to avoid the tedious empirical gain adjustment process that often relies on sets of piecewise constant gains, to improve the convergence speed of the algorithm.