Integrating High Level and Low Level Planning

TL;DR

This paper proposes an integrated planning framework combining high-level global trajectories with low-level robot and crowd planning, improving navigation robustness and enabling real-time operator input, while identifying and addressing failure modes of existing systems.

Contribution

It introduces a probabilistic integrated planning formulation that unifies high and low level decision making, generalizes the ROS navigation stack, and incorporates formal methods for system analysis.

Findings

Generalizes the ROS navigation stack for better robustness

Statistically sound arbitration between planning levels during disturbances

Enables real-time operator input at global and local levels

Abstract

We present a possible method for integrating high level and low level planning. To do so, we introduce the global plan random \emph{trajectory} $\boldsymbol{\eta}_0 \colon [1,T] \to \mathbb R^2$, measured by goals $G_i$ and governed by the distribution $p(\boldsymbol{\eta}_0 \mid \{ G_i\}_{i=1}^m)$. This distribution is combined with the low level robot-crowd planner $p(\mathbf{f}^{R},\mathbf{f}^{1},\ldots,\mathbf{f}^{n}\mid\mathbf{z}_{1:t})$ (from~\cite{trautmanicra2013, trautmaniros}) in the distribution $p(\boldsymbol{\eta}_0,\mathbf{f}^{(R)},\mathbf{f}\mid\mathbf{z}_{1:t})$. We explore this \emph{integrated planning} formulation in three case studies, and in the process find that this formulation 1) generalizes the ROS navigation stack in a practically useful way 2) arbitrates between high and low level decision making in a statistically sound manner when unanticipated local disturbances arise and 3) enables the integration of an onboard operator providing real time input at either the global (e.g., waypoint designation) or local (e.g., joystick) level. Importantly, the integrated planning formulation $p(\boldsymbol{\eta}_0,\mathbf{f}^{(R)},\mathbf{f}\mid\mathbf{z}_{1:t})$ highlights failure modes of the ROS navigation stack (and thus for standard hierarchical planning architectures); these failure modes are resolved by using $p(\boldsymbol{\eta}_0,\mathbf{f}^{(R)},\mathbf{f}\mid\mathbf{z}_{1:t})$. Finally, we conclude with a discussion of how results from formal methods can guide our factorization of $p(\boldsymbol{\eta}_0,\mathbf{f}^{(R)},\mathbf{f}\mid\mathbf{z}_{1:t})$.