CRAM &amp;#x2014; A Cognitive Robot Abstract Machine for everyday manipulation in human environments

Beetz, Michael; Mösenlechner, Lorenz; Tenorth, Moritz

doi:10.1109/iros.2010.5650146

Cited by 159 publications

(111 citation statements)

References 28 publications

Supporting

Mentioning

106

Contrasting

Unclassified

Order By: Relevance

“…Current approaches to learning show that symbol semantics useful for robot manipulation like the qualitative relations used in our representation can also be acquired by means of learning [36]. So far, most systems rely on hand-coded symbols (e.g., [37]) though. On the level of basic spatial relations we also use predefined relations in order to be informed of qualitative reasoning techniques applicable.…”

Section: Discussion Of Related Approachesmentioning

confidence: 99%

Leveraging Qualitative Reasoning to Learning Manipulation Tasks

Wolter

Kirsch

2015

Robotics

View full text Add to dashboard Cite

Learning and planning are powerful AI methods that exhibit complementary strengths. While planning allows goal-directed actions to be computed when a reliable forward model is known, learning allows such models to be obtained autonomously. In this paper we describe how both methods can be combined using an expressive qualitative knowledge representation. We argue that the crucial step in this integration is to employ a representation based on a well-defined semantics. This article proposes the qualitative spatial logic QSL, a representation that combines qualitative abstraction with linear temporal logic, allowing us to represent relevant information about the learning task, possible actions, and their consequences. Doing so, we empower reasoning processes to enhance learning performance beyond the positive effects of learning in abstract state spaces. Proof-of-concept experiments in two simulation environments show that this approach can help to improve learning-based robotics by quicker convergence and leads to more reliable action planning.Keywords: qualitative spatial reasoning; robot learning; AI robotics MotivationRobotic applications have evolved considerably, yet they still lack the efficiency, flexibility and adaptability that is needed to improve our everyday life as versatile service robots. Planning and learning Robotics 2015, 4 254 are the two major paradigms employed to make a robot act intelligently. Planning is a form of symbolic reasoning applied on various levels of abstraction. Planning is effective whenever reliable forward models of the robot and its environment are available. Learning is a powerful method to generate such models from data. However, it requires carefully handcrafted state space representations to converge to high-quality models in reasonable time. Neither planning nor learning on their own suffice to meet the demands of versatile service robots. By integrating both paradigms one can benefit from their respective strengths. So far, a true integration of learning and planning has not been achieved that allows a robot to acquire new skills efficiently in a truly autonomous way.Our work is motivated by this basic question of how to integrate learning and planning, leveraging their individual strengths to produce reliable and versatile robot behavior. We propose a qualitative reasoning framework as the glue between those two paradigms. Qualitative reasoning lifts information, e.g., obtained by perception, to a knowledge level, enabling us to integrate it with common sense knowledge and to use available reasoning techniques [1][2][3]. The qualitative representation serves two purposes: (1) to make a planning or action selection process with learned models more efficient; and (2) to control selection of learning samples in order to make a learning process more reliable and to obtain high-quality models with little manual design effort. We present two experimental studies in simulation for a specific robot task: to throw an object to a specific destination. The results ind...

show abstract

Section: Discussion Of Related Approachesmentioning

confidence: 99%

Leveraging Qualitative Reasoning to Learning Manipulation Tasks

Wolter

Kirsch

2015

Robotics

View full text Add to dashboard Cite

show abstract

“…The approach closest to ours is presented in the works of M. Beetz and his group, where high-level actions are translated, using knowledge-based techniques, into robot programs [28]. However, the resulting code is normally at the level of ROS primitives, acceptable in case of service robots, but without providing any real-time guarantees needed in industrial setting.…”

Section: Related Workmentioning

confidence: 99%

From High-Level Task Descriptions to Executable Robot Code

Stenmark

Malec

Stolt

2015

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

Abstract. For robots to be productive co-workers in the manufacturing industry, it is necessary that their human colleagues can interact with them and instruct them in a simple manner. The goal of our research is to lower the threshold for humans to instruct manipulation tasks, especially sensor-controlled assembly. In our previous work we have presented tools for high-level task instruction, while in this paper we present how these symbolic descriptions of object manipulation are translated into executable code for our hybrid industrial robot controllers.

show abstract

“…To execute these actions the CRAM [18] toolbox based on Common Lisp is adopted. The reasons for using this toolbox are:…”

Section: Executivementioning

confidence: 99%

Integrating planning and execution for ROS enabled service robots using hierarchical action representations

Janssen

Meijl

Marco

et al. 2013

2013 16th International Conference on Advanced Robotics (ICAR)

View full text Add to dashboard Cite

Abstract-The aim of the RoboEarth project is to develop a globally accessible database, that enables service robots to share reusable information relevant to the execution of their daily tasks. Examples of this information are the hierarchical task descriptions, or action recipes, that represent typical household tasks as symbolic action sequences. By annotating these static action representations with hierarchical planner predicates, they can be interpreted by the Hierarchical Task Network planner SHOP2 to compose more flexible, optimized robot plans, based on the actual state of the environment and the available capabilities of the robot. To subsequently execute the composed plans in a typical household environment, the CRAM executive toolbox is adopted, allowing a tight integration between plan execution and run-time knowledge inference. This paper presents the integration of these two components into one cohesive planning and execution framework, tailored for the safe execution of abstract tasks in a challenging household environment. The resulting framework is implemented on the AMIGO service robot and a basic experiment is conducted to demonstrate the frameworks integral functionality.

show abstract

CRAM — A Cognitive Robot Abstract Machine for everyday manipulation in human environments

Cited by 159 publications

References 28 publications

Leveraging Qualitative Reasoning to Learning Manipulation Tasks

Leveraging Qualitative Reasoning to Learning Manipulation Tasks

From High-Level Task Descriptions to Executable Robot Code

Integrating planning and execution for ROS enabled service robots using hierarchical action representations

Contact Info

Product

Resources

About