Motion Control Techniques for Collaborative Multi-Agent Activities

1
Motion Control Techniques for Collaborative
Multi-Agent Activities

• David Benjamin
• Phuoc Nguyen

2
What is an Agent?

• An agent is a system situated in, and part of, an
  environment, which senses that environment, and
  acts on it, over time, in pursuit of its own
  agenda. This agenda evolves from programmed
  goals.
• The agent acts to change the environment and
  influences what it senses at a later time.

3
Motion Control

• In the field of automation, it involves the use
  of devices such as hydraulic pumps, linear
  actuators, or servos to control the position
  and/or velocity of an object.
• In the field of multi-agent, collaborative
  systems it is the control of the position and/or
  velocity of agents so that the agents can work
  together to accomplish a goal.

4
Motion Control
5
Centralized Control

• A single point of control where the controller
  gathers all of the information in the environment
  (including the state of each agent) and the plans
  the motion for the each agent
• Central controller has high level complexity
• Requires a high bandwidth communication link
• May be impractical for battery powered agents

6
Distributed Control

• Each agent determines its motion by sensing the
  environment and then reacting according to a set
  of rules
• Agents are unaware of the agendas of other
  agents.
• Does not require communication with a central
  controller.
• Simpler implementation.
• Flexible.

7
Social Potential Forces

• Initially used for obstacle avoidance.
• Obstacles and agents are assigned negative
  charges
• Goal destinations are assigned positive charges
• A maximum electric field is formed when the agent
  and the obstacles are within close proximity
  (repelling forces).
• A minimum electric field is formed when the agent
  and the obstacles are within close proximity
  (attractive forces).
• The agent will naturally avoid obstacles while it
  moves toward its goal destination.

8
Social Potential Forces
Attractive Force
Repulsive Force
Resultant Field
Agent Path
9
Key Terms

• VLSR - Very Large Scale Robotics System
• Global Controller defines the pair-wise
  potential laws for ordered pairs of components
• Global Control Force resultant force calculated
  by each robot. Global in the sense that it
  coordinates the agents and determines the
  distribution of the agents throughout the system.
• Local Control Force The individual attractive
  and repulsive forces sensed by an agent.
• Leading Agents Mobile agent with a
  preprogrammed path.
• Landmark Agents Have a fixed position. Are
  immune to social potential forces, but imposes
  social potential forces on ordinary agents.
• Ordinary Agents Mobile agent that is subjected
  to social potential forces and also imposes
  social potential forces on other agents.

10
Beehive Simulation

• Each bee is an ordinary agent.
• Imposes a repulsive force on other bees
• Is subjected to attractive forces of the flowers
  and the beehive
• Flowers and beehive are landmark agents.
• Impose attractive and repulsive forces on the bees

11
Potential-Based Implementation

• Agents do not make any decisions
• All movements are triggered by active forces
• All agents implement their own force model
• Flowers and beehive have attractive forces to
  each bee
• Bees have repel force

12
Design

• Simulation class contains main scheduler
• Initialize the scenario
• Control the simulation rate
• Map2D Simulate the environment
• Account for all entities
• Process potential fields
• FlowerAgent
• Represent an area/object of interest
• Supply collectable data (nectar)
• BeehiveAgent
• Represent a sink node
• Store nectar or collectable data
• BeeAgent
• Mobile node that gather nectar
• Move to interest area base on potential fields
  direction and magnitude
• DisplayFrame
• Java base GUI
• Display movement in realtime
• DataCollector

13
Load Balancing

• Mechanism to prevent swarming affect
• Each flower have a queuing service. If queue is
  full, attractive force is greatly reduce
• Attractive force has an inverse distance square
  relationship
• Bees have a repel force on each other
• Bees have a maximum load capacity it can carry
• Force threshold
• As the bee capacity increase, its attraction to
  the hive also increase. And the attraction to
  flowers will decrease. Once hive attraction
  overtake the flower by a certain threshold, the
  bee will change direction and head back to the
  hive.

14
Movement Model
FL
Beehive
FL
FL
15
Simulation Results

• Configuration Four flower with equal nectar
• 10 Bees total
• 2 Exercises, linear and square force model
• Performance is approximately identical

16
Simulation Results

• Configuration Four flower with variable nectar
• 10 Bees total
• 2 Exercises, linear and square force model
• Performance is approximately identical

17
Market-Based Collaboration

• Collaborative mechanism employed by the
  Autonomous Collaborative Mission Systems (ACMS).
• Aimed at controlling groups of heterogeneous
  agents.
• Two stage process
• Bid solicitation
• Contract award

18
Market-Based Collaboration
19
Role-Based Approach

• Based on the E-CARGO model
• Each agent or group agents is described as a
  9-tuple
• ltC,O,A,M,R,E,G,S0gt
• C is a set of classes
• O is a set of objects
• A is a set of agents
• M is a set of messages
• R is a set of roles
• E is a set of environments
• G is a set of groups
• S0 is the initial state of the system

20
Role-Based Approach

• Roles specify how an agent behaves at a specific
  context within a limited period
• Each agent will only respond to a subset of
  messages that are defined by its role.
• Each agent will respond differently to the same
  message based on its role.
• Each agent can be programmed to play many
  different roles based on the state of the
  environment and/or the messages it receives.

21
Demo
22
Questions?
23
Citations