CMAX++ : Leveraging Experience in Planning and Execution using Inaccurate Models
Authors: Anirudh Vemula, J. Andrew Bagnell, Maxim Likhachev6147-6155
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | CMAX++ is also shown to outperform baselines in simulated robotic tasks including 3D mobile robot navigation where the track friction is incorrectly modeled, and a 7D pick-and-place task where the mass of the object is unknown leading to discrepancy between true and modeled dynamics. We test the efficiency of CMAX++ and A-CMAX++ on simulated robotic tasks emphasizing their performance in each repetition of the task, and improvement across repetitions. |
| Researcher Affiliation | Collaboration | Anirudh Vemula1, J. Andrew Bagnell2, Maxim Likhachev1 1 Robotics Institute, Carnegie Mellon University 2 Aurora Innovation |
| Pseudocode | Yes | Algorithm 1 Hybrid Limited-Expansion Search; Algorithm 2 CMAX++ (regular text) and A-CMAX++ (regular and italics text) in small state spaces; Algorithm 3 CMAX++ in large state spaces |
| Open Source Code | No | The paper does not provide an explicit statement or link for open-source code for its methodology. |
| Open Datasets | No | The paper conducts experiments on 'simulated robotic tasks' where 'icy patches placed randomly around the track' and 'random start and obstacle locations' are used. It does not refer to or provide access information for a public, established dataset. |
| Dataset Splits | No | The paper mentions running experiments for a certain number of instances or repetitions (e.g., '10 instances', '20 repetitions') and averaging results, but it does not specify traditional training, validation, or test dataset splits or percentages. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions the 'ss-pybullet library' in the acknowledgements, but it does not provide a specific version number. Other software components are not mentioned with version numbers. |
| Experiment Setup | Yes | For all the approaches, we perform K = 100 expansions. For A-CMAX++, we use a non-increasing sequence with αi = 1 + βi where β1 = 100 and βi is decreased by 2.5 after every 5 repetitions. We chose the metric d as the manhattan metric and use ξ = 0 for this experiment. We use a radius of δ = 3 for the hyperspheres introduced in the 7D discrete state space, and to ensure fair comparison use the same radius for KNN regression. A-CMAX++ uses a non-increasing sequence αi = 1+βi where β1 = 4 and βi+1 = 0.5βi. |