Composition Pattern for Constraint-based Programming with Application to Force-sensorless Robot Tasks

Robot platforms and applications are becoming more autonomous and complex. Many of these robots, and in particular service robots, evolve from a highly structured, human-shielded environments to a less structured, human- populated environments. Moreover, these robots are increasingly expected to interact physically and cognitively with humans and their environment.The overall complexity of these robot systems requires the integration of knowledge of many domains and from multiple experts. Dealing with this integration challenge requires a systematic approach and knowledge-driven, flexible, reusable, and adaptable software.This dissertation makes two complementary contributions: firstly, it provides a systematic approach to deal with the complexity of robot systems in general; and secondly, it provides a way to integrate force-sensorless wrench control in service robots. As its first contribution, this dissertation introduces the Composition Pattern as a uniform and easy-to-grasp systematic approach to deal with complexity, from software architecture to behavior composition. The Composition Pattern defines a Composite Functional Entity as first-class citizen of system design. This Composite Functional Entity separates Coordinator, Composer, Scheduler, Configurator(s), Monitor(s), Communicator(s) and Functional Entities. Each Functional Entity can be in itself a Composite Functional Entity, hence can lead to a hierarchy of interacting entities. The Composition Pattern builds on the metamodeling concept, which considers all entities to be models, not objects. The former forms a constructive way to implement the 5Cs approach of separation of concerns; the latter separates concepts of the domain from implementation details, which makes the models more portable to other robot platforms.Concretely, the Composition Pattern is applied as an architectural pattern to refactor the iTaSC constraint-based programming software framework. This resulted in a framework of more reusable, flexible, robust, and adaptable entities. Furthermore, the Composition Pattern is used to structure and formalize constraint-based programming in a domain-specific language (DSL). This DSL enables developers and users to easier (and hence faster) understand and (re)program constraint-based programming applications, since (i) it provides a template of the system, (ii) it enables automatic model verification, and (iii) it enables manual or automatic code generation to the software framework ecosystem of choice. As its second contribution, this dissertation presents a novel force-sensorless wrench control scheme for velocity controlled robots. This control scheme does not require a precise dynamic model of the robot, environment, or contact point; nor does it require an expensive and complex force sensor. Furthermore, it allows the combination of the wrench control constraints with other constraints, and it features a reference adaptation factor, which can be applied to impose a desired transient behavior on the appliedwrench. Experimental validation of the control scheme, through the integration in the resolved-velocity iTaSC approach and software framework, proves that a stable, constant contact wrench can be reached in a repeatable way, and with an accuracy that fits service robot tasks.In addition, this dissertation integrates both contributions in a force-sensorless and bimanual human-robot comanipulation application. The application shows complex behavior emerging from the composition of constraints to instantaneously and concurrently (i) keep the robot grippers parallel, (ii) null wrenches applied to the grippers, (iii) avoid obstacles, (iv) maintain visual contact with a person, (v) avoid joint limits, and (vi) optimize joint configuration. These constraints are expressed in multiple and different generalized task spaces.The software implementation of the DSL, the iTaSC software framework, the wrench control schemes, and the comanipulation application are all made publicly availabe under an open-source license.Preface Abstract Symbols, Abbreviations and Definitions Contents List of Figures List of Tables 1 Introduction 2 Background and Positioning 3 Systematic Robot Application Development Using the Composition Pattern 4 Constraint-based task modelling and execution using DSL 5 The Composition Pattern Applied to Constraint-Based Programming 6 Force-sensorless Wrench Control 7 Force-Sensorless and Bimanual Human-Robot Comanipulation 8 Conclusions A Proofs B Additional experimental data of force-sensorless force control C Videos Bibliography List of Publicationsnrpages: 285status: publishe