Perception-Based Temporal Logic Planning in Uncertain Semantic Maps

In this article, we address a multi-robot planning problem in environments with partially unknown semantics. The environment is assumed to have a known geometric structure (e.g., walls) and to be occupied by static labeled landmarks with uncertain positions and classes. This modeling approach gives rise to an uncertain semantic map generated by semantic simultaneous localization and mapping algorithms. Our goal is to design control policies for robots equipped with noisy perception systems so that they can accomplish collaborative tasks captured by global temporal logic specifications. To specify missions that account for environmental and perceptual uncertainty, we employ a fragment of linear temporal logic (LTL), called co-safe LTL, defined over perception-based atomic predicates modeling probabilistic satisfaction requirements. The perception-based LTL planning problem gives rise to an optimal control problem, solved by a novel sampling-based algorithm, that generates open-loop control policies that are updated online to adapt to a continuously learned semantic map. We provide extensive experiments to demonstrate the efficiency of the proposed planning architecture.

I. Introduction

Autonomous robot navigation is a fundamental problem that has received considerable research attention [1]. The basic motion planning problem consists of generating robot trajectories that reach a known desired goal region starting from an initial configuration while avoiding known obstacles. More recently, a new class of planning approaches has been developed that can handle a richer class of tasks than the classical point-to-point navigation and can capture temporal and Boolean specifications. Such tasks can be, e.g., sequencing or coverage [2], data gathering [3], or persistent surveillance [4] and can be modeled using formal languages, such as linear temporal logic (LTL) [5], [6]. Several control synthesis methods have been proposed to design correct-by-construction controllers that satisfy temporal logic specifications assuming robots with known dynamics that operate in known environments [7]–[15]. As a result, these methods cannot be applied to scenarios where the environment is initially unknown, and therefore, online replanning may be required as environmental maps are constructed, resulting in limited applicability. To mitigate this issue, learning-based approaches have also been proposed that consider robots with unknown dynamics operating in fully or partially unknown environments. These approaches learn policies that directly map on-board sensor readings to control commands in a trial-and-error fashion; see, e.g., [16]–[19]. However, learning-based approaches, in addition to being data inefficient, are specialized to the environment and the robotic platform they were trained on.