Navigation and guidance of autonomous vehicles is a fundamental problem inrobotics, which has attracted intensive research in recent decades. This reportis mainly concerned with provable collision avoidance of multiple autonomousvehicles operating in unknown cluttered environments, using reactivedecentralized navigation laws, where obstacle information is supplied by somesensor system. Recently, robust and decentralized variants of model predictive control basednavigation systems have been applied to vehicle navigation problems. Propertiessuch as provable collision avoidance under disturbance and provable convergenceto a target have been shown; however these often require significantcomputational and communicative capabilities, and don't consider sensorconstraints, making real time use somewhat difficult. There also seems to beopportunity to develop a better trade-off between tractability, optimality, androbustness. The main contributions of this work are as follows; firstly, the integrationof the robust model predictive control concept with reactive navigationstrategies based on local path planning, which is applied to both holonomic andunicycle vehicle models subjected to acceleration bounds and disturbance;secondly, the extension of model predictive control type methods to situationswhere the information about the obstacle is limited to a discrete ray-basedsensor model, for which provably safe, convergent boundary following can beshown; and thirdly the development of novel constraints allowing decentralizedcoordination of multiple vehicles using a robust model predictive control typeapproach, where a single communication exchange is used per control update,vehicles are allowed to perform planning simultaneously, and coherencyobjectives are avoided.
展开▼
