beat bender

28

Upload: rajan-singh

Post on 25-Dec-2015

87 views

Category:

Documents


3 download

DESCRIPTION

a theory on beat bender

TRANSCRIPT

2

As We proceed

• Beat Bender• Subsumption architecture• Prior Work• System Descriptions• Agents• Rules and Layers• Emergence • Aesthetics

Indian Institute of Information Technology

3

About Beat Bender• BeatBender uses a behavior-based model to elicit emergent

rhythmic output from six autonomous agents

• From an artistic perspective, the rules used to define the agent behavior provide a simple but original composition language.

• From these simple rules emerge unexpected behavioral interactions that direct the formation of complex rhythmic output.

Indian Institute of Information Technology

4

Subsumption Architecture• The use of the subsumption architecture as a means of eliciting

emergent behavior was introduced by “Rodney Brooks “ in his seminal paper “Intelligence without Representation”.

• Avoids explicitly programmed procedural algorithms

• Subsumption architecture systems focus on developing sets of behaviors;

• These behaviors are implemented as layers, with some layers given priority over others

Indian Institute of Information Technology

Layer 3

Layer 2

Layer 1

Layers Subsuming

5

Applicability of SA• It was primarily used in the field of Robotics.

• Applicable to all the fields where an emergent behavior is expected.

• Bryson was first to propose the applicability of SA in the development of a musical component.

Indian Institute of Information Technology

6

Meta-creation and ALS started• Haile an anthromorphic Robotic Percussionist was

first artificial musical meta-creation.

• Kinetic Engine by Eigenfelt used network agent architecture to emulate percussive ensemble.

• Brown explored the rhythm generating potential of cellular automata (CA).

• Tokui and Iba demonstrate the use of interactive evolutionary computation as a means of producing rhythms in their CONGA system .

Indian Institute of Information Technology

7

Why Subsumption Architecture• CA as well as GA can also be applied.

• CA cells are processed in parallel

• Rules are hardcoded inhibiting responsive behavior.

• SA implicitly encode system behavior.

• It nullifies the complexity of GA.

Indian Institute of Information Technology

8

Basic Outline • Autonomous rhythmic sequencer utilizing a scalable multi-agent system.

• We shall proceed using six agents

• Each agent is responsible for the generation of one bit.

• Each Agent can be in two states, ON and OFF. Initially all of them are OFF.

• As the system starts each agent is activated using a set of rules.

• THESE RULES SUBSUME EACH OTHER

Indian Institute of Information Technology

9

Design Considerations• Beat bender project has got two distinct Phases.

• Building the architecture:-

Implementation of a robust tool that allows exploration of various layer rules

This becomes the behavioral basis for the agents.

It provides a flexible architecture for enabling and disabling the rules

Indian Institute of Information Technology

10

Design Considerations contd.

• Explore subsumption rules:-

Exploit subsumption layers and rules in order to elicit emergent behaviors.

This behaviors are evaluated based on the characteristics of the rythmic output.

Indian Institute of Information Technology

11

Beat Bender Agents

• They are capable of triggering ON and OFF audio events.

• The states are computed in parallel.

• SA determines which behavior to be performed based on the current.

• Once the new state has been set the system updates the Environment variables.

• If a particular agent’s state is set to ON then an audio event is triggered and one of three drum sounds plays.

Indian Institute of Information Technology

12

Pictorial representation

Indian Institute of Information Technology

The state of all six agents over three time steps

The Layer Rules Interface tool

13

Analysis Tools

• The BeatBender interface also consists of a graphical representation of the last six patterns played by the system

• The interface also captures each rhythm as a text file.

Indian Institute of Information Technology

Particular agent state over time

One cycle of the systems state

ON

OFF

14

Analysis Tools Continued

• The analysis tools were also essential for evaluating the output produced by the system.

• Each layer of the architecture must be implemented, tested, and modified in order to identify configurations that produce complex agent interactions, and consequently, interesting rhythmic output.

• For BeatBender layer tool allows for the exploration of the various permutations by providing a method for quickly reordering layer rules.

Indian Institute of Information Technology

15

System Components

• Agents

• System Goals

• Behaviors

• Perceptions

• Layers and rules

Indian Institute of Information Technology

BUILDING BLOCKS

16

Agents

• The agent Module acts as a container for the other components.

• Layer Module implements the whole SA.

• Sound Module implements the Audio playback system.

• Timing system is responsible for sequential activition of each agent.

Indian Institute of Information Technology

17

System Goals

• Rhythms produced should not be radically different from one time stamp to another.

• It should slowly evolve while maintaining some regularity.

• The rhythmic output must be aesthetic.

• Quality Criterion to evaluate musicality of each rhythm and its complexity.

Indian Institute of Information Technology

18

System Behaviors

Indian Institute of Information Technology

• Firstly an Agent can change their state i.e. from ON to OFF or FLIP its state.

• Secondly the after processing, an agent is set to the ON state then it also must select a drum sound to use for playback.

• The current implementation of BeatBender provides three sound options, each one a different variation of a tabla drum: a long hit, a solid hit, and a low hit.

19

Agent Perceptions

Indian Institute of Information Technology

• Agent is continually updated with specific perceptions which are used to determine which behaviors are enacted under which conditions.

• Perceptions include the agent’s state, the previous agent’s state, the next agent’s state and the number of currently active agents in the system.

• Each one of these perceptions is used to determine which behavior should be enacted by the system.

• These conditional statements and the behaviors they enact are the basis for the subsumption layers that determine agent behavior.

20

Layers and Rules• Layers are made up of rules . Each rule has an antecedent and a

consequence.

• Each layer is also given a rank rule.

• If the preconditions are met simultaneously only the rules associated with the highest ranked layer will have its behavior enacted.

• The rules in the beat bender can be divided into four categoris. Directed,Temporal,Collectiveand Undirected.

Indian Institute of Information Technology

21

Layers and Rules Contd.• Collective: If there are more than 3 active agents then turn this agent on IF TOTALAGENTS > 3 THEN ON• Directed: If the previous agent is on then flip this agent’s state. IF P = ON THEN FLIP• Temporal: If the agent has been on for the last four cycles then turn it off. IF CONSECUTIVE > 4 THEN OFF• Undirected: Rules are based on information about an agent and its

neighbors but without reference to a specific agent’s state.

Indian Institute of Information Technology

22

Evaluation• Beat bender is evaluated using two criteria :-

• Emergence: The emergence criterion was assessed using simple objective measures.

• Aesthetic: Assessed using subjective measures.

Indian Institute of Information Technology

23

Emergence• The application of individual rules does not produce interesting emergent

behavior

• Only though the interaction of two or more local elements is the process of convergence made visible.

• Individual rules do, however, produce rhythms that reveal two distinct types of behavior :-

Consistent behavior if each agent maintains the same state in every cycle.

Cyclical behavior if the agents’ states alternate on two or more subsequent cycles

Indian Institute of Information Technology

24

Individual RulesRule Applied Pattern Generated Behavior

IF P=OFF then ON X _ X _ X _ ConsistentIF N = OFF then ON X X X X X _ ConsistentIF P=N then FLIP X _ X _ X X

X X X X _ X_ X _ _ _ __ _ X _ X _X X _ _ _ _X X _ X _ _X X X X _ _

Cyclical

IF P = OFF then FLIP X _ X _ X _ _ X X_ _ X _ _ _ X _ _X _ X X _ X X _ __ X X X _ X _ _ __ X X _ X _ X X X_ _ X X X X _ X XX X X _ _ _ _ _ _

Cyclical

Indian Institute of Information Technology

25

Evaluation Example• An example of a pattern exhibiting this type of emergent behavior is generated

using the following set of rules.

• Rule 1: TURN ON AGENT• Rule 2: IF P = N then FLIP• Rule 3: 0 IGNORE, 1 ON, 2 IGNORE, 3 REST• Rule 4: IF STRIKES > 4 then OFF• Rule 5: IF ACTIVEAGENTS > 3 then FLIP

An audio sample of this rhythm can be heard at ww.aaronlevisohn.com/beatbender

Indian Institute of Information Technology

26

Rule Combinations• Using rules in combination with each other facilitates emergent agent

behaviors and the resulting complex rhythmic output.

Indian Institute of Information Technology

#1:IF P = OFF thenON#2:IF N = OFF thenON

X X X X X _ Consistent

#1:IF P = OFF thenFLIP#2:IF N = OFF thenON

X X X _ _ _ X X X X X_ _ _ _ _ X X X X X _

Cyclical

27

References• Aaron Levisohn and Philippe Pasquier,”BeatBender Subsumption Architecture

for Autonomous Rhythm Generation”.

• G. Weinberg and S. Driscoll, “Robot-human interaction with an anthropomorphic percussionist,”.

• A.R. Brown, “Exploring Rhythmic Automata,”Applications of Evolutionary Computing.

• http://www.aaronlevisohn.com/beatbender.html.

Indian Institute of Information Technology

28

Thanks for your patient hearing

Indian Institute of Information Technology