Intermediate Language for Mobile Robots: A link between the high-level planner and low-level services in robots

摘要

The development of service and field robotics has been rapid during the last few decades. New versatile and affordable sensors are now available, and very importantly, computing power has increased very fast. Several intelligent features for robots have been presented. They include the use of artificial intelligence (AI), laser range finders, speech recognition, and image processing. This all has meant that robots can be seen more frequently in ordinary environments, or even in homes. Most development work has concentrated on a single or a few sophisticated features in development projects, but even work to design control structures for different levels in robot control has been done. Several languages for industrial and mobile robots have been introduced since the first robot language WAVE was developed in 1973. Tasks can be given to robots in these languages, but their use is difficult and requires special skills of users. In the future, robots will also be used in homes, and ordinary people should be able to give tasks for robots to perform. This should be done descriptively using natural language as in describing tasks to another person. In this work an intermediate language for mobile robots (ILMR) has been presented. It makes it easier to design a new task for a robot. ILMR is intended for service and field robots and it acts as an intermediate link from user, an intelligent planner or a human-robot interface to a robot's actions and behaviours. The main principle in development work has been simplicity and ease of use. Neither any deep knowledge of robotics nor good programming skills are required when using ILMR. While easy to use, ILMR offers all the required features that are needed to control today's specialised service and field robots. These features contain sequential and concurrent task execution and response to exceptions. ILMR also makes it easier to manage the development of complicated software projects of service robots by creating easy-to-use interfaces to all of several subsystems in robots. It is possible for users to use JLMR to give direct commands or tasks to a robot, but it is intended to be used with higher-level abstract languages, such as sign language or even natural spoken language through a high level planner. An action in ILMR can be given coarsely, i.e. in an abstract way, or in detail. Due to this coarseness, ILMR is suitable to be used with higher-level abstract languages and the set of elementary commands supports directly the use of natural language. With ILMR no complicated models of robots and the world are needed. Only a few measureable parameters for robots are needed and a simple map of the environment is maintained. ILMR has been implemented in two different kinds of robots, and its use and performance has been studied with simulators and actual robots in a wide variety of tests. The structure and operation of ILMR has proved to be useful and several tasks have been carried out successfully using both test robots.