Iterative Temporal Planning in Uncertain Environments With Partial Satisfaction Guarantees

Abstract

This work introduces a motion-planning framework for a hybrid system with general continuous dynamics to satisfy a temporal logic specification consisting of co-safety and safety components in a partially unknown environment. The framework employs a multi-layered synergistic planner to generate trajectories that satisfy the specification and adopts an iterative replanning strategy to deal with unknown obstacles. When the discovery of an obstacle renders the specification unsatisfiable, a division between the constraints in the specification is considered. The co-safety component of the specification is treated as a soft constraint, whose partial satisfaction is allowed, while the safety component is viewed as a hard constraint, whose violation is forbidden. To partially satisfy the co-safety component, inspirations

Managing safety and mission completion via collective run-time adaptation

“…Most of the existing works are based on the assumptions that the environment is static and that each agent has either a global communication range (can communicate with each other agent in the system) or that each agent can obtain a full knowledge of the system and the environment at any time. These assumptions, however, do not usually hold in real-world scenarios [7]. In our work, an adaptation is performed on-the-fly every time an unexpected system or environmental feature is observed in a part of the system.…”

Reactive synthesis for finite tasks under resource constraints

“…There exist several works that consider finite tasks in uncertain environments [16], [17]. These works use a fragment of LTL called co-safe LTL [18] to specify such tasks.…”

Reactive Semantic Planning in Unexplored Semantic Environments Using Deep Perceptual Feedback

“…Unlike the problem of safe navigation in a completely known environment, the setting where the obstacles are not initially known and are incrementally revealed online has so far received little theoretical interest. Some few notable exceptions include considerations of optimality in unknown spaces [16], online modifications to temporal logic specifications [17] or deep learning algorithms [18] that assure safety against obstacles, or the use of trajectory optimization along with offline computed reachable sets [19] for online policy adaptations. However, none of these advances (and, to the best of our knowledge, no work prior to [11]) has achieved simultaneous guarantees of obstacle avoidance and convergence.…”