Synthesis of shared control protocols with provable safety and performance guarantees

Abstract

We formalize synthesis of shared control protocols with correctness guarantees for temporal logic specifications. More specifically, we introduce a modeling formalism in which both a human and an autonomy protocol can issue commands to a robot towards performing a certain task. These commands are blended into a joint input to the robot. The autonomy protocol is synthesized using an abstraction of possible human commands accounting for randomness in decisions caused by factors such as fatigue or incomprehensibility of the problem at hand. The synthesis is designed to ensure that the resulting robot behavior satisfies given safety and performance specifications, e.g., in temporal logic. Our solution is based on nonlinear programming and we address the inherent scalability issue by presenting alternative methods. We assess the feasibility and the scalability of the approach by an experimental evaluation.

Blending Controllers via Multi-Objective Bandits

“…Finally, we review shared control protocols [10,21,15,7,17,20], wherein a robot's commands are blended with its human user. Shared control protocols and blending controllers have the same goal of balancing the safety and performance of their given controllers.…”

“…In [10], the environment dynamics are unknown but assumed to be linear. Finally, the authors in [21] use semi-definite programming for blending controllers, which requires prior knowledge of the environment dynamics and the confidence level of each controller. In contrast to [10] and [21], the approach proposed in this paper is model-free.…”

Robotic assistance in the coordination of patient care

“…Although there is ongoing work trying to develop formal methods for safe human–robot interaction (Jansen et al, 2017; Li et al, 2014), imperfect automation persists, and dependence upon such imperfect automation presents serious problems for robotic assistance during safety-critical tasks. This concern is heightened by results from studies indicating increased trust in and reliance upon embodied systems as compared with virtual or computer-based decision support, suggesting a higher possibility of committing Type I errors.…”

Human-in-the-Loop Mixed-Initiative Control Under Temporal Tasks

“…The goal is to design a shared control policy to satisfy a LTL formula and minimize a cost function. Another recent work [17] addresses the control problem of MDP under LTL formulas where the autonomy strategy is blended into the human strategy in a minimal way that also ensures safety and performance. However, the direct interaction on the low level is not investigated in the aforementioned work.…”