GitHub - radarsat1/qualia: A free (FLOSS) software library for embedded AVR agent-based systems written in simple C++. Licensed under the Gnu GPL.

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 117 Commits
docs		docs
tests		tests
tools/scons		tools/scons
.gitignore		.gitignore
Action.cpp		Action.cpp
Action.h		Action.h
Agent.h		Agent.h
Allocator.cpp		Allocator.cpp
Allocator.h		Allocator.h
Environment.h		Environment.h
INSTALL		INSTALL
LICENSE		LICENSE
NeuralNetwork.cpp		NeuralNetwork.cpp
NeuralNetwork.h		NeuralNetwork.h
Observation.cpp		Observation.cpp
Observation.h		Observation.h
Policy.h		Policy.h
QLearningAgent.cpp		QLearningAgent.cpp
QLearningAgent.h		QLearningAgent.h
QLearningEGreedyPolicy.cpp		QLearningEGreedyPolicy.cpp
QLearningEGreedyPolicy.h		QLearningEGreedyPolicy.h
QLearningPolicy.h		QLearningPolicy.h
QLearningSoftmaxPolicy.cpp		QLearningSoftmaxPolicy.cpp
QLearningSoftmaxPolicy.h		QLearningSoftmaxPolicy.h
Qualia.cpp		Qualia.cpp
Qualia.h		Qualia.h
README		README
RLObservation.cpp		RLObservation.cpp
RLObservation.h		RLObservation.h
RLQualia.cpp		RLQualia.cpp
RLQualia.h		RLQualia.h
Random.h		Random.h
Reward.h		Reward.h
RewardEnvironment.cpp		RewardEnvironment.cpp
RewardEnvironment.h		RewardEnvironment.h
SConstruct		SConstruct
StaticAllocator.cpp		StaticAllocator.cpp
StaticAllocator.h		StaticAllocator.h
avrdefs.cpp		avrdefs.cpp
avrdefs.h		avrdefs.h
build.sh		build.sh
common.h		common.h
error.cpp		error.cpp
error.h		error.h

Repository files navigation

Description
-----------
Qualia is a free software library for embedded AVR agent-based systems written in simple C++.
The software is licensed under the Gnu GPL version 3 (see LICENSE file).

Version
-------
Current version is 0.1.

Authors
-------

Head developer:
Sofian Audry
info(@)sofianaudry(.)com


Thanks
------

This project is realized with the support of CINQ as part of the Emerge research project with 
LabXmodal (Concordia Univeristy, MontrŽal), IDMIL (McGill University, MontrŽal) and 
Moment Factory (MontrŽal).

Objectives
----------

1. Allow to easily switch between different environments
 * Two modes: simulation (computer) vs embedded (AVR).
 * The agent should stay the same.
 * Interchangeable environment simulation vs embedded
 * Easy switch between the two modes

2. Well-managed on AVR chips
 * Low memory trace (Flash and SRAM)
 * Static allocation

3. Modular
 * Agent can stay the same while environments are exchanged
 * eg. easy switch of reward function (reinforcement learning)

4. Multi-paradigm
 * Procedural
 * FSM
 * Reinforcement learning