Evaluation Of Task Realizability For Robot Systems

With the development of technology, robots have enjoyed a widerange of applications in social life. However, there is still a long way tocreate an omnipotent robot. Currently, all robot systems are designed forspecific applications with limited capability of accomplishing tasks.Therefore, it is necessary to evaluate whether a robot system is able tocomplete a prescribed task. For a realizable task, strategy to fulfill the taskis further figured out, while root cause for an unrealizable task is analyzedand strategies should be designed to transform the task to be realizable.Besides, evaluation of task realizability will provide important reference intask allocating process for a multi-robot system. When allocating a task, allrobots which are able to complete the task are firstly selected out and thencompared to find out the best robot under given cost index.This paper focuses on the above issues and main work is as follows:(1) A framework of evaluating task realizability for robot systemsbased on representation spaceThe idea behind the proposed framework is to regard task realizationprocess as a state changing process in a representation space.Representation space model is constructed with state variables which canfully represent the robot system and the given task. The initial and finalstates of the robot system are mapped as the start and goal points in thespace. The space is partitioned into two parts, the reachable andunreachable subspaces according to all constraints the robot system issubject to. The robot system is able to accomplish the task if the followingconditions were satisfied. Firstly, the initial and final states both lie in thereachable subspace. Secondly, a path connecting them can be found within the reachable subspace. The optimal solution to realize the task could befurther figured out by path searching in representation space. Otherwise,causes for an unrealizable task are analyzed and strategies to fix the taskare designed. The critical conditions and the cost to fix the task will beobtained correspondingly. Finally, comparison among different effectivestrategies is conducted to select out the best strategy under the given costindex.(2) Realizability analysis for path planning of a robot manipulatorPath planning of a robot manipulator is investigated in the followingthree circumstances, which are single obstacle, multiple obstacles andvarious constraints. Firstly, the framework is used to help in designingstrategies to fix an unrealizable task and selecting out the best strategy.Secondly, impact on task realizability of different obstacles is analyzedwhich provides guidance in fixing unrealizable tasks. Finally, taskrealizability and strategy design to fix unrealizable tasks are studied undervarious constraints.(3) Parking problem of a wheeled mobile robotThe proposed framework is applied to solve three parking problems,one of which is evaluated as unrealizable since its final state lies inunreachable subspace. The nearest neighbor strategy is proposed totransform the task to be realizable.(4) Configuration design of a modular robotThe configuration design of a modular robot is facilitated with theframework by evaluating the capability to accomplish a task under variousconfigurations.