| Popis: |
One of the most important problems to solve in robotics is the collision avoidance between a robot and its environment. A robot should perceive the risk and have a reactive behaviour before an imminent collision occurs. Path planning is a hard computational problem, so having a fast tool to calculate collisions is a key factor to decrease the necessary time to generate safety trajectories. In applications where no path planning exists, for instance in manual guidance or teleoperation, a real time collision detector is needed so as to avoid collisions and to be able to interact with the environment, for example sliding over a surface. The knowledge of minimum distances between robots or objects that share a workspace enables robots to behave in a predictive way. In the human-robot interaction field, virtual fixtures can be used both to prevent collisions and help the human operator by increasing his performance. In this kind of applications minimum distance and collision detection must be known in real time. A new library: Robotic Proximity Queries (RPQ) package has been developed to deal with these requirements, using PQP as the proximity query engine. The original package has been used to optimize the queries when working with open kinematic chains, like robots. These optimizations have been done with the aim of improving the time performance of the generic package and simplifying its use in robotic environments. A system composed of two robots has been used as a test bed to show the performance of the RPQ library. Two applications that benefit from RPQ performance are presented. First, a robotic assisted surgical application, with an assisted cut of a rigid tissue. The surgeon guides freely the driller held by a slave robotic arm that avoids undesired drillings by means of virtual protections. The second application is a dynamic expansion of the working space by means of multirobot cooperation. With these applications, not only proximity queries are shown, but also the graphical interface and the use of a virtual robot based on RPQ. |
