CSR: Small: Incremental Sampling Methods for Online Reactive Motion Planning with Temporal Logic Specifications

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

The objective of this project is to develop a new class of real-time motion planning and control algorithms, enabling mobile cyber-physical systems such as autonomous cars, aircraft, or mobile robots to behave in a provably safe, reliable and efficient fashion, in dynamically-changing, uncertain environments, shared with humans and other independent agents. Safe interaction of such systems with humans, human-controlled systems, or other automated systems will require the ability to behave according to accepted protocols and rules stated in high-level "human-like" languages, such as those arising from the rules of the road, ethical and privacy concerns, as well as rules of engagement in military or security applications. The core of the project is a new approach to the synthesis of real-time planning and control algorithms for mobile cyber-physical systems, combining incremental sampling algorithms for trajectory generation with efficient local model checking techniques. The proposed approach is intended to provide, in addition to provably correct and safe closed-loop behaviors, both asymptotic optimality guarantees and reactive planning capabilities. As a concrete target application area, autonomous automobiles are considered. In particular, a concept is proposed for autonomy-enabled mobility-on-demand system, which could help reducing traffic congestion in metropolitan areas by providing low-cost, convenient, private, and flexible transportation options to city dwellers.