Mandatory: Big push-button top, 10 USD: https://www.adafruit.com/product/1185 Battery (flo has spares) Chip, 14 USD (worst case flo has a spare): https://www.adafruit.com/product/5395 Case (flo will 3D print) Optional: Audio output (instead of car horn, to play announcement)

0. Abbreviations LL or LLP - Low-level policy, which converts directional commands (linear velocity on the x-y plane and angular velocity around the UP axis) into joint angles 1. Install everything Create a new conda environment and follow the install instructions from the 2 readmes in the subdirectories: https://github.com/simonchamorro/legged-gym-rl You should use Python 3.8. Due to the isaacgym depency, this only works on a machine with a CUDA GPU and at least Nvidia Pascal arch (GTX 1080 or later). You need a specific version of pytorch, as mentioned in the readme.

Base+arm policy We wanted to make a 3-way comparison between different ways of training the base+arm policy: (a) training base and arm separately (b) training base first, then arm (c) training both together <-- This would be the obvious winner For (a, "training base and arm separately"), we want to include the arm when training, but either zero it out or send random actions. Surprisingly, this doesn't work at all. The policy needs control over the arm, otherwise it doesn't learn to compensate for it - both when it's zeroed and when the arm is moving randomly. (There is possibly some bug in here somewhere but the task is very, very similar to the other two and they work fine. Will revisit later.) Training a reaching policy with the arm on its own was easy and leads to better overall grasping range than when the arm is mounted on the pupper but this makes sense because the arm fully stretched out can lead to the Go1 tipping over and terminating the episode (and when training only the arm, the arm is fixed).

Safety :skull_and_crossbones: Stay out of the safety circle when the robot is powered on and deeeefinitely keep the touchscreen out of the safety circle. The robot is powerful enough to punch you in the face and break some bones. :red_circle: Whatever you do, if you control the robot via the touchscreen or you control it via some remote software, the red emergency button always works. Just smack it and the robot will instantly shut down and stop moving. This does not harm the robot in any way, so feel free to use it whenever. Test the button before starting to program the robot. :fast_forward: The arm has built-in safety features that halt the arm if it's going too fast or if it's pushing against something that has too much resistance (like when you set a goal point below the table, it will hit the table but then stop). However, it can still accelerate quickly and it can quickly smash into something before the safety mechanism comes on. So be sure to use it with low max velocity/low acceleration unless you know exactly what you're doing. :electric_plug: The arm is always "stiff", meaning the breaks in each joint are active by default. Don't try to move the arm unless you are HOLDING the "freedrive" touchscreen button or the black little nubbin at the back of the touchpad panel (see below). Only then can you move the arm around, otherwise you damage the breaks. :weight_lifter: The max payload of the robot is 5kg, which is on top of its own mass. Meaning you can pick up and effortlessly move around objects of up to (5kg - weight_of_the_gripper). In order to be allowed to use the arm, put your full name below, indicating that you've read these points and will respect these. Any injuries/damages will be your own responsibility unless it's due to a mechanical defect of the robot.

Belated hohoho! :smiley: Here's everything you need to know about your secret santa present. It's a little thingy that checks your Gmail every half hour or so and simply turns on a light if there's new mails. If you press the big top button (the envelop), the light turns off - basically, you can shush it. I've built the whole thing from scratch (code, electronics, box) and everything is open-source. If you have privacy concerns, I'm linking the code and materials at the end. 0 - Preamble

Overview Big Picture Hardware Perception Planning Control Software Big Picture

Preliminaries Please keep these instructions stricly confidential. Especially do not share this assignment or discuss details of this assignment with people that aren't Krishna Murthy Jatavallabhula, Florian Golemo, or Liam Paull while solving the assignment and up until three months after receiving the assignment. This assignment is intended for individual evaluation and not as team work. Overview In short, there are 3 parts and you have 6 full days to solve all of these: You have to train up a policy on an RL environment You have to write up a quick (1-2p max) summary on what you did and why You have to make a super short presentation for a paper of our choice (15min, ideally with a few slides)

(A) Use IEEE PDF eXpress to validate your paper for inclusion in IEEE Xplore There is a strict size limit of 10 MB for PDF files in the conference proceedings. The page limit for all workshop papers is 6 (not including references). Papers longer than 6 pages will not be processed and will not appear on IEEE Xplore. IEEE requires that all papers to be entered into the IEEE Xplore digital library meet the IEEE style guidelines. LaTeX and MS-Word style files compatible with IEEE requirements are available from the CVPR web site: http://cvpr2020.thecvf.com/submission/main-conference/author-guidelines Papers that do not meet these guidelines will not be placed in Xplore. This is IEEE policy. Note: PDF eXpress does not collect the papers. After validating your paper, you must still upload it to the workshop’s camera-ready paper collection website (CMT) for production, as described in Step 2. There are two ways to use IEEE PDF eXpress. If you generate your own PDF file, PDF eXpress can validate it. If you are using any of the following document formatting systems, PDF eXpress can generate an Xplore-compliant document from the following source files: MS-Word, WordPerfect, RTF (rich text format), Freelance, LaTeX, PageMaker, FrameMaker, Word Pro, or Quark. If you used LaTeX, PageMaker, or Quark to prepare your paper, there are a few extra steps required to supply the images to IEEE. Login to PDF express: http://www.pdf-express.org. The conference ID is: 50498X

# Nepal Winter School 2019, RL lab assignment ## Setup ### ONLY ON WINDOWS: Go to https://www.microsoft.com/en-us/download/details.aspx?id=100593, download the program and install it. Also have Anaconda Python installed in version 3.6 or 3.7. You also need to install some other packages for python: conda install -c anaconda mpi4py conda install -c conda-forge pystan conda install -c conda-forge pyglet conda install -c conda-forge swig pip install Box2D # to use box2d ph