Mobile Robot Lab Home

EAGER: CNS: Misdirection in Robot Teams: Exploiting Organizational Principles for Operational Advantage

Directions to the Lab

Overview | People | Publications | Software | Multimedia | Links
  Overview

Trust, dependability, cohesion, and capability are integral to an effective team. These attributes are the same for teams of collaborating robots and humans as well as all-robot teams. For a team to remain effective in the presence of faults, failures, and errors from nature or intent, team members must have sufficient perception and understanding of the problem and possible solutions to make appropriate adjustments in their operations. When multiple teams with competing incentives are tasked, a strategy, if available, may be to weaken, influence or sway the attributes of other teams and limit their understanding of their full range of options. Such strategies are widely found, for example, in nature, sporting contests, and military operations, where strategies such as feints and misdirection often appear. This project focuses on one class of higher-level strategies for multi-robots: namely, to misdirect and where feasible to counter misdirection. As multi-robot systems become more autonomous, distributed, networked, numerous, and with more capability to make critical decisions, the prospect for intentional and unintentional misdirection must be anticipated. This project will support robust multi-robotic operations in important strategic domains such as security and military operations.

The project studies strategies to enable co-robots, multi-robots and teams of multi-robots to model, generate, and cope with misdirection in various situations. This research direction in robotic control offers a novel approach to resilience in and among these teams to these forms of possible disruption. Computational models, drawn particularly from studies of human endeavors and group behaviors, provide a general framework for understanding, producing, and countering misdirection in robotic systems. A framework of computational models will be designed using recursive schema-theoretic models of behaviors at the individual and team levels, building on decentralized methods of control and communication, to provide robustness in the presence of noisy and chaotic environments.

 
  People


 
  Publications

  • Push and Pull: Shepherding Multi-Agent Robot Teams in Adversarial Situations
  • M. J. Pettinati, R. C. Arkin
    pdf (4.8 MB)

  • Wolves in Sheep’s Clothing: Using Shill Agents to Misdirect Multi-Robot Teams
  • M. J. Pettinati, R. C. Arkin, and A. Krishnan
    pdf (3 MB)

  • Misdirection in Robot Teams: Methods and Ethical Considerations
  • R. C. Arkin
    pdf (127 KB)


 
  Software


 
  Multimedia

    We are trying to misdirect a group of agents (move them from one location to another). Agents are influenced by either a pushing (repulsive) agent alone or a pushing agent and a pulling (shill) agent. Each mark agent (the agent to be moved) has a threshold. This threshold is the number of agents the mark needs to be able to see moving with the deceptive agents or urging it to move to follow along. Each marks begins responding to the pushing and pulling agents when its individual threshold has been reached and stops responding when below the threshold.

    Physical Platforms

    In the case with no shills, the group splits. The repulsive robot is able to successfully push one of the mark agents to the end (this agent must have a threshold of one). The other two agents, however, remain outside of the goal location. In the case with one shill agent, all agents reach there threshold and are pulled by the shill and pushed by the repulsive agent to the goal location.
    With Shill Agent (14 MB)
    No Shill Agent (42.1 MB)

    Moreover, we present a counter-misdirection approach for the shill-based misdirection by developing a new type of agent: counter-misdirection agents (CMAs). These CMAs are able to detect the misdirection process and stop marks that are misdirected by shill agents and its leader from reaching the goal location.
    Counter-Misdirection Agents (33.3 MB)

    Simulation Results In Environments With Obstacles

    Simulation showing the group of marks agents being moved from the target to the goal in an environment with an object. The repulsive agent is orange. The mark agents are yellow when below their thresholds and green when they are above their thresholds. The shill is brown.
    Simulation Video (7 MB)


 
  Links

  • Forthcoming