Robust behavior and perception using hierarchical state machines: a pallet manipulation experiment

Abstract

Abstract-Interacting with simple objects in semi-controlled environments is a rich source of challenging situations for mobile robots, particularly when performing sequential tasks. In this paper we present the computational architecture and results obtained from a pallet manipulation experiment with a real robot. To achieve a good success rate in locating and picking the pallets a set of behaviors is assembled in a hierarchical state machine. The behaviors are arranged in such a way that the global uncertainty of the task is progressively reduced when approaching the goal. To do so, actions are generated in each stage that increase the confidence of the robot of being in that particular relation to the world. In order to set up this experiment, it is required a non-trivial set of working senso-motor behaviors. We build on this set to design and test a pallet moving task in which the robot has to locate, approach, obtain the pose, pick up and, finally move the pallet to its target position. The only sensory sources of information available to the robot are a binocular vision system and its internal odometry. To carry out this task we have equipped a RobEx robot with a 1 DOF forklift and a 4 DOF binocular head. We present the conceptual and computational models and the results of the experiments in a real setup.

A Perception-aware Architecture for Autonomous Robots

“…In the absence of the proper tools, robots tend to be programmed using hard-coded if-then-else constructs (which are considerably error-prone when the complexity increases) or, at most, fixed plans embedded in state machines (see [34,35]). Using state machines to embed plans makes code more structured, easier to understand and less-likely to contain programming errors in comparison with hard-coded logic.…”

Recognition and Location Algorithm for Pallets in Warehouses Using RGB-D Sensor

“…In [19,20], the pose of single pallets was estimated based on geometry features extracted from distance data provided by LiDAR with geometry classifiers. In addition, the authors of [21][22][23][24][25][26] obtained pallet pose information based on pallet size and edge features extracted from a color image provided by a camera. To reduce the effect of the environment, pallets were identified with marks attached to the pallet feet in [27][28][29].…”