Full-general motion planning under nonlinear dynamics and arbitrary uncertainty distributions

Develop a motion planning methodology for autonomous systems operating in dynamic environments that rigorously accounts for uncertainty in the motion of other agents in full generality, specifically for nonlinear system dynamics and arbitrary uncertainty distributions, avoiding simplifying assumptions such as linear dynamics or Gaussian/bounded uncertainty models.

Background

The paper addresses safe motion planning among dynamic agents when the agents’ future motion follows an unknown distribution and must be quantified online. Many existing approaches rely on simplified assumptions—such as linear system dynamics and Gaussian or bounded uncertainty—to make uncertainty handling tractable.

In the Introduction, the authors explicitly state that extending motion planning approaches to handle uncertainty without these simplifying assumptions, i.e., for nonlinear dynamics and arbitrary distributions, remains unresolved. Their adaptive conformal prediction framework is presented as a step toward addressing uncertainty adaptively, but it does not fully solve the general case the authors identify as open.