Symbolic reasoners generate plans which are often not exploiting the robot capabilities and are sensitive to runtime disturbances. This work proposes a scheduler as an interface between a discrete, symbolic plan and a motion control based on constraint optimization. Acting as a local reasoner, the scheduler valuates a set of predicates to decide when an action will be executed. Given a task specification which describes how the action should be realized, the scheduler configures the controller at runtime. A demonstration will be provided considering an “open drawer” scenario

No abstract

Abstract-Task specification models define the activities to be executed by a robot in order to achieve its goal. Classical examples are the sequences involved in assembly or pick and place tasks. This work introduces the preview coordination execution model, an extension to the traditional way in which the execution of such task specifications is coordinated at runtime. Instead of taking activities one-by-one as defined in the task specification model, preview coordination optimizes the task scheduling based on knowledge about the likelihood that not just the activities required by the current state can be executed, but that also one or more of those related to future states of the system can be activated. An experiment with mobile manipulation tasks illustrates the benefits of preview coordination.

Abstract- T HE problem of robotic task definition and execution was pioneered by Mason, [1], who defined setpoint constraints where the position, velocity, and/or forces are expressed in one particular task frame for a 6-DOF robot. Later extensions generalized this approach to constraints in i) multiple frames, ii) redundant robots, iii) other sensor spaces such as cameras, and iv) trajectory tracking. Our work extends tasks definition to i) expressions of constraints, with a focus on expressions between geometric entities (distances and angles), in place of explicit set-point constraints, ii) a systematic composition of constraints, iii) runtime monitoring of all constraints (that allows for runtime sequencing of constraint sets via, for example, a Finite State Machine), and iv) formal task descriptions, that can be used by symbolic reasoners to plan and analyse tasks. This means that tasks are seen as ordered groups of constraints to be achieved by the robot's motion controller, possibly with different set of geometric expressions to measure outputs which are not controlled, but are relevant to assess the task evolution. Those monitored expressions may result in events that trigger switching to another ordered group of constraints to execute and monitor.For these task specifications, formal language definitions are introduced in the JSON-schema modeling language.