Adaptive Trajectory Estimation with Power Limited Steering Model under Perturbation Compensation
Trajectory estimation of maneuvering objects is applied in numerous tasks like navigation, path planning and visual tracking. Many previous works get impressive results in the strictly controlled condition with accurate prior statistics and dedicated dynamic model for certain object. But in challenging conditions without dedicated dynamic model and precise prior statistics, the performance of these methods significantly declines. To solve the problem, a dynamic model called the power-limited steering model (PLS) is proposed to describe the motion of non-cooperative object. It is a natural combination of instantaneous power and instantaneous angular velocity, which relies on the nonlinearity instead of the state switching probability to achieve switching of states. And the renormalization group is introduced to compensate the nonlinear effect of perturbation in PLS model. For robust and efficient trajectory estimation, an adaptive trajectory estimation (AdaTE) algorithm is proposed. By updating the statistics and truncation time online, it corrects the estimation error caused by biased prior statistics and observation drift, while reducing the computational complexity lower than O(n). The experiment of trajectory estimation demonstrates the convergence of AdaTE, and the better robust to the biased prior statistics and the observation drift compared with EKF, UKF and sparse MAP. Other experiments demonstrate through slight modification, AdaTE can also be applied to local navigation in random obstacle environment, and trajectory optimization in visual tracking.