GitPlanet

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details
All Projects → Sohojoe → ActiveRagdollControllers

Sohojoe / ActiveRagdollControllers

Licence: Apache-2.0 license
Research into controllers for 2d and 3d Active Ragdolls (using MujocoUnity+ml_agents)

Programming Languages

C#
18002 projects
python
139335 projects - #7 most used programming language
Jupyter Notebook
11667 projects

Labels

video-gamesreinforcement-learningdeep-learningunityunity3ddeep-reinforcement-learningopenaineural-networksdeepmindml-agents

Projects that are alternatives of or similar to ActiveRagdollControllers

DQN-using-PyTorch-and-ML-Agents
A simple example of how to implement vector based DQN using PyTorch and a ML-Agents environment
Stars: ✭ 81 (+170%)
Mutual labels:  deep-reinforcement-learning, ml-agents
reinforce-js
[INACTIVE] A collection of various machine learning solver. The library is an object-oriented approach (baked with Typescript) and tries to deliver simplified interfaces that make using the algorithms pretty simple.
Stars: ✭ 20 (-33.33%)
Mutual labels:  deep-reinforcement-learning, deepmind
Pytorch-RL-CPP
A Repository with C++ implementations of Reinforcement Learning Algorithms (Pytorch)
Stars: ✭ 73 (+143.33%)
Mutual labels:  openai, deepmind
jax-rl
JAX implementations of core Deep RL algorithms
Stars: ✭ 61 (+103.33%)
Mutual labels:  deep-reinforcement-learning, deepmind
Ml In Tf
Get started with Machine Learning in TensorFlow with a selection of good reads and implemented examples!
Stars: ✭ 45 (+50%)
Mutual labels:  deep-reinforcement-learning, deepmind
Reinforcement Learning
Learn Deep Reinforcement Learning in 60 days! Lectures & Code in Python. Reinforcement Learning + Deep Learning
Stars: ✭ 3,329 (+10996.67%)
Mutual labels:  deep-reinforcement-learning, deepmind
ddrl
Deep Developmental Reinforcement Learning
Stars: ✭ 27 (-10%)
Mutual labels:  deep-reinforcement-learning, openai
Advanced Deep Learning And Reinforcement Learning Deepmind
🎮 Advanced Deep Learning and Reinforcement Learning at UCL & DeepMind | YouTube videos 👉
Stars: ✭ 121 (+303.33%)
Mutual labels:  deep-reinforcement-learning, deepmind
Pysc2 Examples
StarCraft II - pysc2 Deep Reinforcement Learning Examples
Stars: ✭ 722 (+2306.67%)
Mutual labels:  deep-reinforcement-learning, deepmind
Deep Reinforcement Learning
Repo for the Deep Reinforcement Learning Nanodegree program
Stars: ✭ 4,012 (+13273.33%)
Mutual labels:  deep-reinforcement-learning, ml-agents
Mujocounity
Reproducing MuJoCo benchmarks in a modern, commercial game /physics engine (Unity + PhysX).
Stars: ✭ 47 (+56.67%)
Mutual labels:  deep-reinforcement-learning, deepmind
Learning To Communicate Pytorch
Learning to Communicate with Deep Multi-Agent Reinforcement Learning in PyTorch
Stars: ✭ 236 (+686.67%)
Mutual labels:  deep-reinforcement-learning, deepmind
Reinforcement-Learning-on-google-colab
Reinforcement Learning algorithm's using google-colab
Stars: ✭ 33 (+10%)
Mutual labels:  deep-reinforcement-learning
awesome-rl
Awesome RL: Papers, Books, Codes, Benchmarks
Stars: ✭ 105 (+250%)
Mutual labels:  deep-reinforcement-learning
abc
SeqGAN but with more bells and whistles
Stars: ✭ 25 (-16.67%)
Mutual labels:  deep-reinforcement-learning
awesome-codex
A list dedicated to products, demos and articles related to 🤖 OpenAI's Codex.
Stars: ✭ 115 (+283.33%)
Mutual labels:  openai
Deep-Reinforcement-Learning-With-Python
Master classic RL, deep RL, distributional RL, inverse RL, and more using OpenAI Gym and TensorFlow with extensive Math
Stars: ✭ 222 (+640%)
Mutual labels:  deep-reinforcement-learning
king-pong
Deep Reinforcement Learning Pong Agent, King Pong, he's the best
Stars: ✭ 23 (-23.33%)
Mutual labels:  deep-reinforcement-learning
yarll
Combining deep learning and reinforcement learning.
Stars: ✭ 84 (+180%)
Mutual labels:  deep-reinforcement-learning
AI-Projects
AI项目（强化学习、深度学习、计算机视觉、推荐系统、自然语言处理、机器导航、医学影像处理）
Stars: ✭ 39 (+30%)
Mutual labels:  deep-reinforcement-learning
View All Similar Projects ➔

ActiveRagdollControllers

Research into controllers for 2d and 3d Active Ragdolls (using MujocoUnity+ml_agents)

Contributors

Download builds (Mac, Windows): see Releases

Controller004

Controller004

  • Type: Discrete 2D
  • Build (MacOS): v0.004 Playable
  • Actions: No-op, Forward, Backwards, Jump, Jump+Forward, Jump+Backwards
  • Controls: Left arrow, Right arrow, Spacebar
  • Mujoco Model: DeepMindHopper
  • Hypostheis: Use discreate + random trainer to create human controller.
  • Outcome:
    • SUCCESS - contriol feels responsive
    • ... Has emerging functionality - i.e. tap left for small step, swap direction in air

Controller003

Controller003

  • Type: Continuous 2D
  • Build (MacOS): v0.003
  • Actions: Forward / Backwards
  • Mujoco Model: DeepMindHopper
  • Hypostheis: Use an adversarial hierarchical trained agent as the controller which gets the inverse reward of the locomation agent on a slower time step. The idea is that it will push the locomoation agent to focus on its weakest areas.
  • Outcome:
    • FAIL - training is too heavily influenced by the number of steps the controller agent takes between decisions;
    • ... it maybe better to train a seperate agent on hyper-parms (i.e. meta learning)
    • ... having read more about these approaches (MAML, RL2, etc) it would be better to move to a Discreate conrtroller as ml-agents LSTM does not work well with Continuous actions.

Controller002

Controller002

  • Type: Continuous 2D
  • Build (MacOS, Windows): v0.002
  • Actions: Forward / Backwards
  • Input: Unity Axis input (left/right or a/d or joystick)
  • Mujoco Model: DeepMindHopper
  • Hypostheis: Use ml-agent (player + Heuristic) and unity input to create a player contoller .
  • Outcome:
    • Works - makes a simple 2d proof of concept of 2d active ragdoll using RL.
    • I needed to use Discrete target velocities for stable training (-1,1,0) however it is continuous under player control.
    • Training was sensitive to the number of steps between chaning the input (see counter logic below)
  • Notes:
    • Controller002InputBrain
      • BrainType: Player = input is from player
      • BrainType: Heuristic = input is from Controller002InputBrain.cs
    • Controller002InputDecision.cs -
      • Waits until counter is 0 then:
      • Takes random action and outputs AxisX as -1, 0, 1 (used as target velocity), then:
      • Set counter random 40-240
    • Controller002InputAgent.cs -

Controller001

Controller001

  • Type: Discrete 2D
  • Build (MacOS): v0.001
  • Actions: Forward / Backwards
  • Mujoco Model: DeepMindHopper
  • Hypostheis: It should be simple to train a backwards / forwards by giving the agent a +1 / 1 velocity target which feeds the reward function.
  • Outcome: It worked - but was harder than expected to find good hyper-parms. Next, wrap this in a proper ml-agent controller and allow user input.
Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].