Learning and Reasoning for Robot Sequential Decision Making under Uncertainty
Authors: Saeid Amiri, Mohammad Shokrolah Shirazi, Shiqi Zhang2726-2733
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In experiments, a mobile robot is tasked with estimating human intentions using their motion trajectories, declarative contextual knowledge, and human-robot interaction (dialog-based and motion-based). Results suggest that, in efficiency and accuracy, our framework performs better than its no-learning and no-reasoning counterparts in office environment. |
| Researcher Affiliation | Academia | 1SUNY Binghamton, Binghamton, NY 13902 USA 2University of Indianapolis, Indianapolis, IN 44227 USA samiri1@binghamton.edu; shirazim@uindy.edu; szhang@cs.binghamton.edu |
| Pseudocode | Yes | We present LCORPP in the form of Algorithms 1 and 2, where Algorithm 1 calls the other. |
| Open Source Code | No | The paper does not provide an explicit statement or link to the open-source code for the LCORPP framework or its components described in the paper. A demo video link is provided, but no code repository. |
| Open Datasets | Yes | We use a human motion dataset (Kato, Kanda, and Ishiguro 2015) to train the LSTM-based classifier, where the dataset was collected using multiple 3D range sensors mounted overhead in a shopping center environment. |
| Dataset Splits | Yes | The data was split into sets for training (70%) and testing (30%). |
| Hardware Specification | No | The paper does not provide specific hardware details such as CPU/GPU models, memory, or processing units used for training or running the experiments. |
| Software Dependencies | No | The paper mentions software like Keras and off-the-shelf POMDP solvers, but does not provide specific version numbers for these software components or programming languages, which are necessary for full reproducibility. |
| Experiment Setup | Yes | Our LSTM s hidden layer includes 50 memory units. In order to output binary classification results, we use a dense layer with sigmoid activation function in the output layer. We use Adam (Kingma and Ba 2014), a first-order gradient method, for optimization. The loss function was calculated using binary cross-entropy. For regularization, we use a dropout value of 0.2. The memory units and the hidden states of the LSTM are initialized to zero. The epoch size (number of passes over the entire training data) is 300. The batch size is 32. |