Learning Concepts from Sensor Data of a Mobile Robot

Klingspor, Volker; Morik, Katharina; Rieger, Anke D.

doi:10.1007/978-1-4613-0471-5_8

Cited by 6 publications

(5 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Bridging the representational gap between the continuous sensorimotor world of a robotic system with the discrete symbols and rules has been a key research goal from the early days of intelligent robotics (Kuipers et al, 2017;Murphy & Murphy, 2000). While grounding predefined symbols in the sensorimotor experience of the robot has been widely used for intelligent robot control (Klingspor et al, 1996;Petrick et al, 2008;Mourao et al, 2008;Wörgötter et al, 2009;Kulick et al, 2013), some argue that symbols "are not formed in isolation", and "they are formed in relation to the experience of agents" (Sun, 2000). We share this viewpoint that has been investigated in a number of studies.…”

Section: Related Workmentioning

confidence: 99%

DeepSym: Deep Symbol Generation and Rule Learning for Planning from Unsupervised Robot Interaction

Ahmetoğlu

Seker

Piater

et al. 2022

jair

View full text Add to dashboard Cite

Symbolic planning and reasoning are powerful tools for robots tackling complex tasks. However, the need to manually design the symbols restrict their applicability, especially for robots that are expected to act in open-ended environments. Therefore symbol formation and rule extraction should be considered part of robot learning, which, when done properly, will offer scalability, flexibility, and robustness. Towards this goal, we propose a novel general method that finds action-grounded, discrete object and effect categories and builds probabilistic rules over them for non-trivial action planning. Our robot interacts with objects using an initial action repertoire that is assumed to be acquired earlier and observes the effects it can create in the environment. To form action-grounded object, effect, and relational categories, we employ a binary bottleneck layer in a predictive, deep encoderdecoder network that takes the image of the scene and the action applied as input, and generates the resulting effects in the scene in pixel coordinates. After learning, the binary latent vector represents action-driven object categories based on the interaction experience of the robot. To distill the knowledge represented by the neural network into rules useful for symbolic reasoning, a decision tree is trained to reproduce its decoder function. Probabilistic rules are extracted from the decision paths of the tree and are represented in the Probabilistic Planning Domain Definition Language (PPDDL), allowing off-the-shelf planners to operate on the knowledge extracted from the sensorimotor experience of the robot. The deployment of the proposed approach for a simulated robotic manipulator enabled the discovery of discrete representations of object properties such as ‘rollable’ and ‘insertable’. In turn, the use of these representations as symbols allowed the generation of effective plans for achieving goals, such as building towers of the desired height, demonstrating the effectiveness of the approach for multi-step object manipulation. Finally, we demonstrate that the system is not only restricted to the robotics domain by assessing its applicability to the MNIST 8-puzzle domain in which learned symbols allow for the generation of plans that move the empty tile into any given position.

show abstract

Section: Related Workmentioning

confidence: 99%

DeepSym: Deep Symbol Generation and Rule Learning for Planning from Unsupervised Robot Interaction

Ahmetoğlu

Seker

Piater

et al. 2022

jair

View full text Add to dashboard Cite

show abstract

“…Our approach uses a representation that is close to the real inputs and outputs of system, with that intermediate type of planning between high-level and reactive planning. Other integrated planning and learning systems for robotic tasks are [Bennet and DeJong, 1996] and [Klingspor et al, 1996]. The first one deals with the concept of permissiveness, that defines qualitative behavior for the operators.…”

Section: Related Workmentioning

confidence: 99%

Advances in Artificial Intelligence - IBERAMIA-SBIA 2006

Sichman¹,

Coelho²,

Rezende³

2006

Lecture Notes in Computer Science

View full text Add to dashboard Cite

the 9th Brazilian Neural Networks Symposium). This decision was a consequence of the successful event organized in 2000, when the First International Joint Conference IBERAMIA/ SBIA 2000 (7th Ibero-American Artificial Intelligence Conference and 15th Brazilian Artificial Intelligence Symposium) occurred in Brazil. Moreover, in 2006 the artificial intelligence community celebrated the golden anniversary of the 1956 Dartmouth Conference that marked the beginning of artificial intelligence as a research field.SBIA 2006 was the 18 th conference of the SBIA conference series, which is the leading Brazilian conference for the presentation of AI research and applications. Since 1995, SBIA has become an international conference, with papers written in English, an international Program Committee, and proceedings published in Springer's Lecture Notes in Artificial Intelligence (LNAI) series.IBERAMIA 2006 was the 10 th conference of the IBERAMIA conference series, which has been one of the most suitable forums for Ibero-American AI researchers (from South and Central American countries, Mexico, Spain and Portugal) to present their results. Following the SBIA and EPIA (Portuguese Conference on AI) experiences, from IBERAMIA 1998 on, it has also become an international conference with proceedings published in Springer's LNAI series.The International Joint Conference was held in Ribeirão Preto, Brazil, during October 23-27, 2006. The call for papers was very successful. In the registration phase, when the authors were asked to submit just the title and abstract of their papers, 281 submissions were received. In the second phase, 246 authors uploaded their full paper to be reviewed. These submissions came from 23 different countries 1 , as shown in Table 1. After the revision process, 62 papers were accepted to be published in these proceedings.In order to improve the overall quality of this joint conference, a double-blind reviewing procedure was adopted, and the acceptance rate was fixed around 25% of the submitted papers. A large group of reviewers from all content areas was set up to carry out the difficult and challenging work of selecting the papers. Every paper was evaluated by at least three reviewers, and ranked by the JEMS system. Whenever conflicts were found, open discussions among the reviewers were triggered and analyzed later to support our final decision.The conference included keynote speeches, introductory and advanced tutorials by world-leading researchers, several workshops covering specific topics and a thesis and dissertation contest.

show abstract

“…In particular, many of the learned clauses find patterns that simply do not make sense from a temporal perspective and, in turn, generalize poorly. We believe a reasonable alternative to our approach may be to incorporate syntactic biases into ILP systems as done, for example, in Cohen (1994), Dehaspe and De Raedt (1996), Klingspor, Morik, and Rieger (1996). In this work, however, we chose to work directly in a temporal logic representation.…”

Section: Learning Temporal Patternsmentioning

confidence: 99%

Specific-to-General Learning for Temporal Events with Application to Learning Event Definitions from Video

Fern¹,

Givan²,

Siskind³

2002

jair

View full text Add to dashboard Cite

We develop, analyze, and evaluate a novel, supervised, specific-to-general learner for a simple temporal logic and use the resulting algorithm to learn visual event definitions from video sequences. First, we introduce a simple, propositional, temporal, event-description language called AMA that is sufficiently expressive to represent many events yet sufficiently restrictive to support learning. We then give algorithms, along with lower and upper complexity bounds, for the subsumption and generalization problems for AMA formulas. We present a positive-examples-only specific-to-general learning method based on these algorithms. We also present a polynomialtime-computable "syntactic" subsumption test that implies semantic subsumption without being equivalent to it. A generalization algorithm based on syntactic subsumption can be used in place of semantic generalization to improve the asymptotic complexity of the resulting learning algorithm. Finally, we apply this algorithm to the task of learning relational event definitions from video and show that it yields definitions that are competitive with hand-coded ones.1. In our formal analysis, we will use two different notions of generality (semantic and syntactic). In this section, we ignore such distinctions. We note, however, that the algorithm we informally describe later in this section is based on the syntactic notion of generality. each timeline of © ¼ subsumes ©. This implies © © ¼ . ¾We can now characterize the AMA LGG using IS and IG.Theorem 14. IG´Ë ©¾¦ IS´©µµ is an AMA LGG of the set ¦ of AMA formulas.Proof: Let ¦ © ½ © Ò and © ½ ¡ ¡ ¡ © Ò . We know that the AMA LGG of ¦ must subsume , or it would fail to subsume one of the © . Using "and-to-or" we can represent as a disjunction of MA timelines given by ´Ï IS´© ½ µµ ¡ ¡ ¡ ´Ï IS´© Ò µµ. Any AMALGG must be a least-general formula that subsumes -i.e., an AMA LGG of the set of MA timelines Ë IS´©µ © ¾ ¦ . Theorem 13 tells us that an LGG of these timelines is given by IG´Ë IS´©µ © ¾ ¦ µ. ¾ 9. There must be at least one such timeline, the timeline where the only state is ØÖÙ © ½ ℄ syn © ¾ ℄µ. Using this property, it is straightforward to show how to compute the syntactic AMA LGG using the syntactic AMA LGG algorithm.Proposition 28. For any AMA formulas © ½ © Ñ , let © be the syntactic AMA LGG of © ½ ℄ © Ñ ℄ . Then, ½ ©℄ is the unique syntactic AMA LGG of © ½ © Ñ .Proof: We know that for each , © ℄ syn ©-thus, since ½ preserves syntactic subsumption, we have that for each , © syn ½ ©℄. This shows that ½ ©℄ is a generalization of the inputs.We now show that ½ ©℄ is the least such formula. For the sake of contradiction assume that ½ ©℄ is not least. It follows that there must be a © ¼ ¾ AMA such that © ¼ ×ÝÒ ½ ©℄ and for each , © syn © ¼ . Combining this with the fact that preserves syntactic subsumption, we get that © ¼ ℄ ×ÝÒ © and for each , © ℄ © ¼ ℄. But this contradicts the fact that © is an LGG; so we must have that ½ ©℄ is a syntactic AMA LGG. As argued elsewhere, the uniqueness of this LGG follows from the f...

show abstract

Learning Concepts from Sensor Data of a Mobile Robot

Cited by 6 publications

References 6 publications

DeepSym: Deep Symbol Generation and Rule Learning for Planning from Unsupervised Robot Interaction

DeepSym: Deep Symbol Generation and Rule Learning for Planning from Unsupervised Robot Interaction

Advances in Artificial Intelligence - IBERAMIA-SBIA 2006

Specific-to-General Learning for Temporal Events with Application to Learning Event Definitions from Video

Contact Info

Product

Resources

About