A collision-free motion planning method by integrating complexity-reduction SLAM and learning-based artificial force design

In order to generally deal with the rotor-type UAV's collision-free motion planning problem in the unknown static environment, we propose a non-holonomic solution via integration of the KF-based SLAM technique and governing force design. The traditional SLAM is modified and reduced as a low-complexity form according to the fact that too early detected obstacle information can be regarded as nearly frozen after sufficient correction. The artificial force terms are designed in a intuitive and smart way, through employment of the wall-following rule and lessons from historical and current experience, which are taught by the bat's predation process. Further, they are converted to the real-time thrust vector expectation. Multiple simulation tests in both continuous and discrete scenes indicate that: (1) using slight sacrifice on the state estimate covariance can exchange pronounced reduction on structural complexity of the complete SLAM in return; (2) the LBAFD can not only mitigate limitations on the path oscillation, no passage between closely spaced obstacles and goal unreachability, but also lead to a high flying and exploration efficiency; (3) the integrated method demonstrates a relatively stable performance under different parameter settings and is even unconcerned to the surrounding characteristics.