Multi-Robot Formation. Dr. Daisy Tang

Transcription

1 Multi-Robot Formation Dr. Daisy Tang

2 Objectives Understand key issues in formationkeeping Understand various formation studied by Balch and Arkin and their pros/cons Understand local vs. global control Be able to determine best formation for given circumstances

3 Formation-Keeping Objective: Robots maintain specific formation while collectively moving along path Examples: Column formation Line formation

4 Key Issues in Formations What is the desired formation? How do robots determine their desired positionsin the formation? How do robots determine their actual positionsin the formation? How do robots move to ensure that a formation is maintained? What should robots do if there are obstacles? How do we evaluate robot formation performance?

5 What is Desired Formation? Dependent upon environment: Obstacle-free A few obstacles Cluttered Dependent upon sensing/communication capabilities and requirements: Ability to detect other robot positions Ability to communicate with each other Ability to sense effect of formation-keeping through the world

6 Possible Formations Formations can be hard-coded, in the sense that they specify Cartesian positions for all robots.

7 Possible Formations (con t.) Or, formations can be defined by constraints, which allow variation in Cartesian positions for robots Move object across room Maintain line-of-sight visibility

8 Some Formation Control Videos

9 Behavior-Based Formation Control for Multi-Robot Teams Presented by Iain Lee

10 About This Paper Author: Tucker Balch, Member, IEEE Ronald C. Arkin, Senior Member, IEEE Published in: IEEE Transactions on Robotics and Automation December, 1998

11 Introduction Why formation-keeping? Formations allow individual team members to concentrate their sensors across a portionof the environment, while their partners cover the rest It is important when sensor assets are limited Widely used in military applications Potential applications in robotics: robot scout, search and rescue, agricultural coverage tasks, and security patrols Formation-keeping objective: Robots maintain specific formation while moving along path

12 Four Major Formation Type Line, column, diamond and wedge Used by U.S. Army mechanized scout platoons on the battlefield

13 Formation Maintenance Two steps: Detect-formation-position(perceptual) Maintain-formation(motor) Three techniques for formation position determination Unit-center-referenced Leader-referenced Neighbor-referenced

14 Motor Schema-Based Formation Control Basic motor schemas (behaviors): move-to-goal avoid-static obstacle avoid-robot maintain-formation noise Each schema generates a vector representing the desired behavioral response A gain valueis used to indicate the relative importance of the individual behaviors

15 Recall the Motor Schema Approach

16 Motor Schema Parameters High-level combined behavior is generated by multiplying the outputs of each primitive behavior by its gain, then summing and normalizing the results.

17 Maintain-Formation Vector Direction: Always in the direction of the desired formation position Magnitude: Depends on how far the robot is away from the desired position Length of arrow = m = magnitude Angle of arrow = d = direction

18 When There Are Obstacles To avoid obstacles like barriers, choices are: Move as a unit around the barrier Divide into subgroups Depends on the relative strengths of behaviors(gain)

19 So, What s the Result? Qualitative/Quantitative View

20 Experiment Setup Georgia Tech s MissionLabrobot simulation Sensors allow a robot to distinguish between robots, obstacles and goals Robots have several navigation waypoints to follow

21 A Qualitative Analysis (1) Four robots in formations

22 A Qualitative Analysis (2) Move around obstacles and through turns Leader-referenced (left) vs. unit-center-referenced (right)

23 A Quantitative Analysis (1) Three metrics: path ratio, position error and timeout of formation Two experiments: a 90 degrees turn and an obstacle field

24 A Quantitative Analysis (2) Results: For 90-degree turn: Diamond formation best with unit-center-ref. Wedge, line formations best with leader-ref. For obstacle field: Column formation best with either unit-centerref. or leader-ref. Most of the time: Unit-center-ref. formations perform better than leader-ref. formations

25 But Unit-center-ref. formations are not widely used: If using human leader, leader-ref. is better For communications restricted applications, the unit-center-ref. requires a transmitter and receiver for each robot and a protocol Passive sensors are difficult to use for unit-center-ref.

26 Conclusion This paper presents a behavior-based approach to implement formationkeeping, several different formation types are implemented and compared But there are some issues that are not discussed in this paper, like scalability and implementation of other possible formation types

27 Designing Control Laws for Cooperative Agent Teams Presented by Raymond Luc

28 Local vs. Global Control Local control laws: No robot has all pertinent information Appealing because of their simplicityand potential to generate globally emergent functionality But, may be difficult to design to achieve desired group behavior Global control laws: Centralized controller (or all robots) possess all pertinent information Generally allow more coherent cooperation But, usually increase inter-agent communication

29 Global Control Use global goals of the team and/or global knowledge about the team s current or upcoming actions to direct an individual s actions How do agents obtain such global knowledge? Centralized informant Interpretation through agent modeling Shortcomings: No adequate global info available Agent does not use all global info provided Processing requires time and resources, hard for real-world applications Changing global info

30 Local Control Guide actions based on information derived from agent s sensory data Allows agent to react to dynamics Global functionality can emergefrom interaction of local control laws However, certain global goals cannot be achieved through the use of local control laws alone

31 Tradeoffs between Global and Local Control Question #1: How static is the global knowledge? In general, the more static the global knowledge is, the more practical its use by a global control law Question #2: How difficult it is to approximate global knowledge? Question #3: How badly will performance degrade without the use of global knowledge? How difficult is it to use global knowledge? In general, the more unknown the global information is, the more dependence a team must have on local control, combined with approximations to global knowledge based on observation

32 Conflicting Global and Local Controls Global control laws use strictly global information Local control laws use strictly local information A better way: view global information as providing general guidance for longer-term actions, whereas local information indicates more short-term, reactive actions A composite control law

33 Formation Keeping Assumptions Each agent has ability to sense locations of its neighbors relative to itself (local knowledge) Agent is physically constrained by inability to move backwards Global goal: Agent reaches destination asap Agents maintain formation Global knowledge: Path the group to take (waypoints) Path the leader is taking

34 Demonstration of Tradeoffs in Formation-Keeping Performance is measured by cumulative formation error: Strategies to investigate: Local control alone Local control + global goal Local control + global goal + partial global knowledge Local control + global goal + more complete global knowledge

35 Formation-Keeping Objective

36 Strategy I: Local Control Group leader knows path waypoints Each robot assigned local leader+ position offset from local leader As group leader moves, individual robots maintain relative positionto local leaders

37 Strategy II: Local Control + Global Goal Group leaderknows path waypoints Each robot assigned global leader+ position offset from global leader As group leader moves, individual robots maintain relative positionto global leader

38 Strategy III: Local Control + Global Goal + Partial Global Knowledge Group leaderknows path waypoints Each robot assigned global leader+ position offset from global leader Each robot knows next waypoint As group leader moves, individual robots maintain relative positionto global leader

39 Strategy IV: Local Control + Global Goal + More Complete Global Knowledge Group leaderknows path waypoints Each robot assigned global leader+ position offset from global leader Each robot knows current and next waypoints As group leader moves, individual maintain relative positionto global leader

40 Time and Cumulative Formation Error

41 Summary of The Formation Case Study Important to achieve proper balance between local and global knowledge and goals Static global knowledge easy to use as global control law Local knowledge appropriate when can approximate global knowledge Local control information should be used to ground global knowledge in the current situation