Learning Visual Servoing with Deep Features and Fitted Q-Iteration

Authors: Alex X. Lee, Sergey Levine, Pieter Abbeel

ICLR 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the performance of the model for visual servoing in a simulated environment.
Researcher Affiliation Collaboration UC Berkeley, Department of Electrical Engineering and Computer Sciences Open AI International Computer Science Institute {alexlee gk,svlevine,pabbeel}@cs.berkeley.edu
Pseudocode Yes Algorithm 1 FQI with initialization of policy-independent parameters
Open Source Code Yes The environment for the synthetic car following benchmark is available online as the package City Sim3D1, and the code to reproduce our method and experiments is also available online2.
Open Datasets Yes Our choice of semantic features are derived from the VGG-16 network (Simonyan & Zisserman, 2014), which is a convolutional neural network trained for large-scale image recognition on the Image Net dataset (Deng et al., 2009).
Dataset Splits Yes We ran the overall algorithm for only S = 2 sampling iterations and chose the parameters that achieved the best performance on 10 validation trajectories.
Hardware Specification No The paper does not specify the hardware (e.g., CPU, GPU models, memory) used for running the experiments. It only mentions a simulated environment.
Software Dependencies No The paper mentions software like Caffe and ADAM, but does not provide specific version numbers for the key software dependencies required for reproducibility.
Experiment Setup Yes The dynamics models were trained with ADAM using 10000 iterations, a batch size of 32, a learning rate of 0.001, and momentums of 0.9 and 0.999, and a weight decay of 0.0005.