Transparent, Flexible, and Resource-adaptive Image Processing for Autonomous Service Robots

摘要

. This paper describes SRIPPs, structured reactive image processing plans, that are tailored for the needs of autonomous service robots. SRIPPs are implemented on top of a modular, dynamically configurable, architecture called RECIPE and specified as subplans written in the robot control/plan language RPL [13]. Specifying image processing routines transparently and explicitly as part of the robots' plans rather than hiding them in separate modules makes the robots' visual capabilities more effective, efficient, and robust. The use of SRIPPs enables robots to generate, reason about and revise their image processing routines. They can also synchronize their image processing activities with their other actions and use control structures to recover from image processing failures. 1 Introduction Many autonomous service robots that act in dynamically changing environments need to acquire and interpret various kinds of visual information. Consider an office delivery robot that is to accompl...