Multi-fingered Hand Grasps with Visuo-Tactile Fusion via Multi-Agent Deep Reinforcement Learning

Authors: Peida Jia, Xuanheng Li, Tianqiang Zhu, Rina Wu, Xiangbo Lin, Yi Sun

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The grasping results on 8 objects show that our approach can achieve stable and compliant grasps. To the best of our knowledge, this is the first work that employs a finger-based multi-agent reinforcement learning approach to control the dexterous grasping process under the guidance of both visual and tactile feedback.
Researcher Affiliation Academia School of Information and Communication Engineering, Dalian University of Technology, China EMAIL, EMAIL
Pseudocode No The paper describes the methodology in text and provides figures (Figure 2 and Figure 3) illustrating the system architecture and network of the grasping phase. However, there are no explicit pseudocode blocks or algorithms labeled as such.
Open Source Code No The paper does not contain any explicit statement about making the source code available, nor does it provide a link to a code repository or supplementary materials for code.
Open Datasets Yes The objective of this work is to effectively integrate visual and tactile information using MADRL methods to achieve stable dexterous grasps with only a single-frame reference grasp. This reference can be obtained from recent static grasp synthesis methods (Zhu et al. 2021) or a grasp dataset (Wang et al. 2023).
Dataset Splits No The paper mentions that experiments are evaluated 'across 50 trials for each object' and that 'All methods are trained under the same experimental settings, including reward functions, iteration counts, and static reference grasps.' However, it does not specify explicit training/test/validation dataset splits with percentages or sample counts for the data used in the reinforcement learning setup.
Hardware Specification Yes The hardware required for the experiments includes an Intel Core T M i9-10920 CPU @3.50GHz 24, and an NVIDIA Geforce RTX 3090 graphics card.
Software Dependencies No We verify our VT-MAGIC scheme using the Mu Jo Co physics simulator. This mentions a software component but does not specify a version number.
Experiment Setup No The paper describes reward functions (with weighting coefficients α1, α2, βi, α3, α5, α6, α7 but without specific values) and termination conditions for the approaching and grasping phases. It also mentions 'iteration counts' and 'preset threshold' for epochs but does not provide concrete numerical values for hyperparameters such as learning rates, batch sizes, or the actual number of epochs/iterations.