Lars Carøe Sørensen

Bio: Lars Carøe Sørensen is an academic researcher from University of Southern Denmark. The author has contributed to research in topics: Kernel density estimation & Robot. The author has an hindex of 4, co-authored 14 publications receiving 54 citations. Previous affiliations of Lars Carøe Sørensen include Maersk.

Optimisation of Trap Design for Vibratory Bowl Feeders

Abstract: Vibratory bowl feeders (VBFs) are a widely used option for industrial part feeding, but their design is still largely manual. A subtask of VBF design is determining an optimal parameter set for the passive devices, called traps, which the VBF uses to ensure correct part orientation. This paper proposes a fast and robust strategy for optimising traps, which makes use of dynamic simulation to efficiently evaluate the performance of parameter sets. The optimisation strategy is based on Bayesian Optimisation and selects new parameter sets to evaluate, using a modified Upper Confidence Bound with regression by Kernel Density Estimation as function estimator. The optimisation is run for four different traps with an industrial part and the best parameter sets are tested for robustness in simulation. The traps are then combined to create two sequences performing orientation of the parts and the designs are prototyped and tested on a real VBF.

Applying Simulation and a Domain-Specific Language for an Adaptive Action Library

Abstract: In this paper, we present the status of ongoing research aimed at tackling the issues of programming robots for small-size productions where fast set-up times, quick changeovers and easy adjustments are essential. We use a probabilistic approach where uncertainties are taken into account, making the deterministic requirements of an assembly process less strict. Concretely, actions from an action library are modelled through parameters, simulation is used to facilitate learning of uncertainty-tolerant actions, and a Domain-Specific Language (DSL) is used to convert the abstractly specified actions into corresponding executable actions. The approach is tested on an application example from industry.

Towards Digital Twins for Industrial Assembly - Improving Robot Solutions by Intuitive User Guidance and Robot Programming

Abstract: Simulation of robotic tasks allows for cheap evaluation and process optimization, which can then be transferred to the physical system. To avoid discrepancies between physical execution and simulation, real-world process data can be fed back to the simulation environment, a concept referred to as a "Digital Twin". This requires the development of a software architecture, that supports Digital Twins for robot tasks.In this work, we propose a system where an operator can take apart a complex assembly, thus creating digitized assembly instructions. These instructions are then used to visually program the robot setup by blocks, which contain functionality ranging from point-to-point motions to high-level skills. These "Skillblocks" allow for a seamless transition between execution in the simulation environment and on the physical robot through interchangeable execution layers in the software architecture. The system also allows for feedback from a physical execution to be monitored in real-time and fed back to the simulation environment for processing.The aim of the system is to close the gap between digital and physical workcells when integrating robot solutions. This increases intuitiveness and allows for process monitoring and optimization through direct feedback to the digital model.

Online Action Learning using Kernel Density Estimation for Quick Discovery of Good Parameters for Peg-in-Hole Insertion

Abstract: Learning action parameters is becoming an ever more important topic in industrial assembly with tendencies towards smaller batch sizes, more required flexibility and process uncertainties. This paper presents a statistical online learning method capable of handling these issues. The method uses elimination of unpromising parameter sets to reduce the elements of the discretised sample space (inspired by Action Elimination) based on regression uncertainty. Kernel Density Estimation and Wilson Score are explored as internal representations. Based on a dynamic simulator setup for a real world Peg-in-Hole problem, it is shown that the presented method can drastically reduce the number of samples needed. Furthermore, it is also shown that the solution obtained in simulation by our learning method succeeds when executed on the corresponding real world setup.

Towards robot cell matrices for agile production–SDU Robotics' assembly cell at the WRC 2018

Abstract: To support shifting to high mix/low volume production, manufacturers in high wage countries aim for robotizing their production operations – with a special focus on the late production phases, wher...

A review of machine learning for the optimization of production processes

Abstract: Due to the advances in the digitalization process of the manufacturing industry and the resulting available data, there is tremendous progress and large interest in integrating machine learning and optimization methods on the shop floor in order to improve production processes. Additionally, a shortage of resources leads to increasing acceptance of new approaches, such as machine learning to save energy, time, and resources, and avoid waste. After describing possible occurring data types in the manufacturing world, this study covers the majority of relevant literature from 2008 to 2018 dealing with machine learning and optimization approaches for product quality or process improvement in the manufacturing industry. The review shows that there is hardly any correlation between the used data, the amount of data, the machine learning algorithms, the used optimizers, and the respective problem from the production. The detailed correlations between these criteria and the recent progress made in this area as well as the issues that are still unsolved are discussed in this paper.

A Survey on Domain-specific Modeling and Languages in Robotics

Abstract: The development of advanced robotic systems is challenging as expertise from multiple domains needs to be integrated conceptually and technically. Model-driven engineering promises an efficient and flexible approach for developing robotics applications that copes with this challenge. Domain-specific modeling allows to describe robotics concerns with concepts and notations closer to the respective problem domain. This raises the level of abstraction and results in models that are easier to understand and validate. Furthermore, model-driven engineering allows to increase the level of automation, e.g. through code generation, and to bridge the gap between modeling and implementation. The anticipated results are improved efficiency and quality of the robotics systems engineering process. Within this contribution, we survey the available literature on domain-specific modeling and languages that target core robotics concerns. In total 137 publications were identified that comply with a set of defined criteria, which we consider essential for contributions in this field. With the presented survey, we provide an overview on the state-of-the-art of domain-specific modeling approaches in robotics. The surveyed publications are investigated from the perspective of users and developers of model-based approaches in robotics along a set of quantitative and qualitative research questions. The presented quantitative analysis clearly indicates the rising popularity of applying domain-specific modeling approaches to robotics in the academic community. Beyond this statistical analysis, we map the selected publications to a defined set of robotics subdomains and typical development phases in robotic systems engineering as reference for potential users. Furthermore, we analyze these contributions from a language engineering viewpoint and discuss aspects such as the methods and tools used for their implementation as well as their documentation status, platform integration, typical use cases and the evaluation strategies used for validation of the proposed approaches. Finally, we conclude with recommendations for discussion in the model-driven engineering and robotics community based on the insights gained in this survey.

Teaching a Robot the Semantics of Assembly Tasks

Abstract: We present a three-level cognitive system in a learning by demonstration context. The system allows for learning and transfer on the sensorimotor level as well as the planning level. The fundamentally different data structures associated with these two levels are connected by an efficient mid-level representation based on so-called “semantic event chains.” We describe details of the representations and quantify the effect of the associated learning procedures for each level under different amounts of noise. Moreover, we demonstrate the performance of the overall system by three demonstrations that have been performed at a project review. The described system has a technical readiness level (TRL) of 4, which in an ongoing follow-up project will be raised to TRL 6.

Automatic error recovery in robot assembly operations using reverse execution

Abstract: Robotic assembly tasks are in general difficult to program and require a high degree of precision. As the complexity of the task increases it becomes increasingly unlikely that tasks can always be executed without errors. Preventing errors beyond a certain point is economically infeasible, in particular for small-batch productions. As an alternative, we propose a system for automatically handling certain classes of errors instead of preventing them. Specifically, we show that many operations can be automatically reversed. Errors can be handled through automatic reverse execution of the control program to a safe point, from which forward execution can be resumed. This paper describes the principles behind automatic reversal of robotic assembly operations, and experimentally demonstrates the use of a domain-specific language that supports automatic error handling through reverse execution. Our contribution represents the first experimental demonstration of reversible computing principles applied to industrial robotics.