A Framework for Integrating Symbolic and Sub-Symbolic Representations
Authors: Keith Clark, Bernhard Hengst, Maurice Pagnucco, David Rajaratnam, Peter Robinson, Claude Sammut, Michael Thielscher
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | An instantiation is developed for a real robot building towers of blocks, subject to human interference; this hierarchy uses a node with a concurrent multitasking teleo-reactive program, a node embedding a physics simulator to provide spatial knowledge, and nodes for sensor processing and robot control. We instantiate the architecture with a Baxter robot tasked to build multiple towers. We conducted several experiments tasking Baxter to build towers of blocks starting from different initial configurations. This implementation was evaluated empirically to determine the robustness and limitations of the implementation at each level of the hierarchy. |
| Researcher Affiliation | Academia | Keith Clark Imperial College Bernhard Hengst University of NSW Maurice Pagnucco University of NSW David Rajaratnam University of NSW Peter Robinson University of Queensland Claude Sammut University of NSW Michael Thielscher University of NSW |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code. It links to a video demonstrating the robot's operation but not to code: 'A video of Baxter building the tower LAB on the right table and COT on the left table, shows the TR program coping with both help and hindrance from a human.2http://robocup.web.cse.unsw.edu.au/baxter/20160127LABCOT-HIx4.mp4' |
| Open Datasets | No | The paper describes experiments with a 'Baxter robot tasked to build multiple towers' using 'uniquely labelled blocks' in a 'blocks-world' environment. This indicates a custom experimental setup rather than the use of a publicly available dataset with concrete access information. |
| Dataset Splits | No | The paper does not specify training, validation, or test dataset splits in the conventional machine learning sense. The system operates in a real-world environment with sensing and action updates. |
| Hardware Specification | Yes | We instantiate our architectural framework with an off-the-shelf Baxter robot tasked to build towers of uniquely labelled blocks. Baxter’s two arms can be operated concurrently and have an overlapping work-space. Each arm end-effector is equipped with a two-pronged gripper and a camera that point down at the table as shown in Figure 1. There are three tables. Baxter’s sensor information includes the arm position and arm effort measurements, 2D camera images, and gripper states. |
| Software Dependencies | No | The framework is instantiated in ROS using five nodes... We employ Open CV to find squares representing blocks and to identify block symbols. The spatial node employs the Gazebo physics simulator... The paper mentions software names (ROS, Open CV, Gazebo) but does not provide specific version numbers for these components. |
| Experiment Setup | No | The paper describes the components of the system (Controller nodes, Spatial node, Symbolic TR node) and their functionalities, including concepts like visual feedback control and use of arm effort in feedback loops. It mentions 'fixed camera offsets' as part of determining block positions. However, it does not provide specific hyperparameters (e.g., learning rates, batch sizes) or system-level training settings in a way that allows direct replication of a training process, as it is a real-robot system. |