Handling Unforeseen Failures Using Argumentation-Based Learning

Abstract

General Purpose Service Robots operate in different environments of a dynamic nature. Even the robot's programmer cannot predict what kind of failure conditions a robot may confront in its lifetime. Therefore, general purpose service robots need to efficiently handle unforeseen failure conditions. This requires the capability of handling unforeseen failures while the robot is performing a task. Existing research typically offers special-purpose solutions depending on what has been foreseen at the design time. In this research, we propose a general purpose argumentation-based architecture which is able to autonomously recover from unforeseen failures. We compare the proposed method with existing incremental online learning methods in the literature. The results show that the proposed argumentation-based learning approach is capable of learning complex scenarios faster with a lower number of observations. Moreover, the final precision of the proposed method is higher than other methods.

Argue to Learn: Accelerated Argumentation-Based Learning

“…This research is an expansion of our previous paper [1], [2]. The specific expansions are listed as follows.…”

“…Using the combination of AF and BAF, argumentation-based learning has been proven to outperform state-of-the-art online incremental learning methods [1], [2]. ABL extracts a set of relevant hypotheses from the learning instances in an online manner and explicitly represents the knowledge acquired from the learning instances as an explainable set of rules as arguments and defeasibilty relations among them.…”

“…A recent study on the comparison of the state-of-the-art methods for incremental online machine learning [15] shows that Incremental Support Vector Machines (ISVM) [16], [17] together with LASVM [18], which is an online approximate SVM solver, and Online Random Forest (ORF) [19] outperform the other methods. The original ABL approach outperformed all these methods in terms of accuracy and learning speed [1], [2]. Therefore, we only compare the proposed AABL method with the original ABL approach.…”

“…Argumentation-Based Learning (ABL) [1], [2] outperformed other online incremental learning approaches and was shown to be successful for handling unforeseen failures. However, ABL is not efficient in terms of space and computational complexity.…”

Local-HDP: Interactive Open-Ended 3D Object Categorization in Real-Time Robotic Scenarios

“…However, in real-life robotic scenarios, a robot can always face new object categories while operating in the environment. Therefore, the model should get updated in an open-ended manner without completely retraining the model [7]. Furthermore, object category recognition is not a well-defined problem because of the large intra-category variation (Figure 1 (a)), multiple object views for each object (Figure 1 (b)), and concept drift in dynamic environments [8].…”