Navigation and path planning using reinforcement learning for a Roomba robot

摘要

Service robots are expected to help us in different tasks in various environments such as homes, hospitals and offices. This work presents the first steps towards building a service robot at our university. The robot is provided with a topological map of a building floor (environmental map). Using this map the robot learns a path from one location to another by means of reinfocerment learning. During execution the robot is provided with a navigation map, related to the environmental map. We make use of a comercial Roomba robot and show how to create an interface in order to control it. Service robots are expected to help us in different tasks in various environments such as homes, hospitals and offices. This work presents the first steps towards building a service robot at our university. The robot is provided with a topological map of a building floor (environmental map). Using this map the robot learns a path from one location to another by means of reinfocerment learning. During execution the robot is provided with a navigation map, related to the environmental map. We make use of a comercial Roomba robot and show how to create an interface in order to control it.