Science Robotics

The PDF file includes:

• Supplement S1. Synopsis of findings.
• Supplement S2. Materials and methods.
• Supplement S3. Experiments and analysis.
• Fig. S1. Three sensorimotor tasks used in our experiments.
• Fig. S2. Different computer vision modalities used in our experiments, illustrated on the urban driving task.
• Fig. S3. Performance of agents equipped with different input representations.
• Fig. S4. Performance of agents equipped with different predicted input representations.
• Fig. S5. Performance of unsupervised, predicted, and ground-truth vision agents as a function of training set size.
• Fig. S6. Performance of image-only and vision-equipped agents as a function of training set size on battle.
• Fig. S7. Imitation learning agent architecture.
• Fig. S8. Network architecture used to infer the vision modalities.
• Table S1. Performance of image-only and vision-equipped agents on training tracks.
• References (11–15

)

Download PDF

Other Supplementary Material for this manuscript includes the following:

• Movie S1 (.mp4 format). Summary.