Automated Planning Techniques for Robot Manipulation Tasks Involving Articulated Objects

Abstract

The goal-oriented manipulation of articulated objects plays an important role in real-world robot tasks. Current approaches typically pose a number of simplifying assumptions to reason upon how to obtain an articulated object's goal configuration, and exploit ad hoc algorithms. The consequence is twofold: firstly, it is difficult to generalise obtained solutions (in terms of actions a robot can execute) to different target object's configurations and, in a broad sense, to different object's physical characteristics; secondly, the representation and the reasoning layers are tightly coupled and interdependent. In this paper we investigate the use of automated planning techniques for dealing with articulated objects manipulation tasks. Such techniques allow for a clear separation between knowledge and reasoning, as advocated in Knowledge Engineering. We introduce two PDDL formulations of the task, which rely on conceptually different representations of the orientation of the objects. Experiments involving several planners and increasing size objects demonstrate the effectiveness of the proposed models, and confirm its exploitability when embedded in a real-world robot software architecture.

An ASP-Based Framework for the Manipulation of Articulated Objects Using Dual-Arm Robots

“…In [5,6] a similar framework based on automated reasoning methodologies has been presented. Such framework employs PDDL language and automated planning engines for the planning module, and Description Logic (DL) solvers in the configuration module, where data are explicitly stored in an ontology, while we use a uniform language and approach (ASP-based) in the whole framework.…”

“…Such framework employs PDDL language and automated planning engines for the planning module, and Description Logic (DL) solvers in the configuration module, where data are explicitly stored in an ontology, while we use a uniform language and approach (ASP-based) in the whole framework. Moreover, differently from most of our approaches, encodings and solvers employed in [5,6] are not currently able to return shortest plans, which is otherwise important, given that in this context executing the actions can be expensive. In [19], instead, a custom-designed multi-robot platform is presented, focused on HRI in indoor service robot for understanding natural language requests.…”

“…Attention has been put to the development of approaches and algorithms for generating the sequence of movements a robot has to perform in order to manipulate an articulated object. In the literature, the problem of determining the 2D configuration of articulated or flexible objects has received much attention in the past few years [5,6,23,27], whereas the problem of obtaining a target configuration via manipulation has been explored in motion planning [2,26,28]. A limitation of such manipulation strategies is that they are often crafted specifically for the problem at hand, with the relevant characteristics of the object and robot capabilities being either hard coded or assumed; thus, in these contexts generalisation property and scalability are somehow limited.…”

Collaborative Robotic Manipulation: A Use Case of Articulated Objects in Three-dimensions with Gravity

“…The problem of determining the 2D or 3D configuration of articulated (or flexible) objects has originated much research work in the past few years [6], [7], [8], [9], [10], and different strategies have been employed as far as motion planning is concerned [11], [12], [13]. Conceptually speaking, the outcome of such motion planning approaches is a continuous mapping in joint space from an initial to a target object's configuration, subject to a number of simplifying hypotheses [12], which lead to two open challenges: (i) sequences of manipulation actions are to be robust with respect to errors in perception and execution, as well as to unpredictable actions carried out by human operators in human-robot collaboration (HRC) scenarios, thereby focusing on generalisation and scalability [7], and (ii) physical laws must be taken into account at the planning level, e.g., the effect of gravity on the object configuration, for predicting its evolution (and therefore estimate) over time [9].…”

Automated Planning Encodings for the Manipulation of Articulated Objects in 3D with Gravity

“…The manipulation of articulated objects plays an important role in real-world robot tasks, both in home and industrial environments [18,15]. In literature, the problem of determining the two-or three-dimensional (2D or 3D) configuration of articulated or flexible objects has received much attention in the past few years [28,21,6,7], whereas the problem of obtaining a target configuration via manipulation has been explored in motion planning [30,3,26]. However, the employed manipulation strategies are often crafted specifically for the problem at hand, with the relevant characteristics of the object and robot capabilities being either hard coded or assumed, thus undermining generalisation and scalability.…”