Basic concepts for human robot shared authority

1
Basic concepts for human / robot shared
authority
HCP 2008, Delft, The Netherlands
Stéphane Mercier, Catherine Tessier, Frédéric
Dehais
www.onera.fr
2
Outline

• Context, State of the Art and Objectives
• Three Main Functions
• Towards a First Formalization
• Experimental Setup
• Conclusion

3
Context

• A mission operated in the physical world by
  several agents, including one or several humans
• An Authority Sharing problem, with its
  dynamicity
• between artificial agents (automatic pilot, UAVs,
  robots...)
• between artificial agents and human operators
• Example of missions
• Monitoring
• Observations
• Search and Rescue

4
Authority Sharing example

• Mission Fire fighting
• A fireman supervising a robot
• To detect fires
• To extinguish starting fires

Buildings
?
5
State of the Art

• Adjustable autonomy
• Autonomy a relationship between entities about
  an object Castelfranchi 2003
• Autonomy ability of the agent to reduce the need
  of operator supervision Goodrich 2001
• Automation levels Sheridan 1978
• Autonomy measure Huang 2005

6
State of the Art

• Conclusion
• Autonomy levels are unsatisfying predefined, not
  operational or mission specific
• Human operator always considered as a perfect
  solution, whereas she/he is fallible
• Need of a formal basis for authority sharing who
  does what, based on which objective criteria?

7
Objectives

• Why, when and how should an agent take the
  initiative?
• When the environment has changed?
• When an agent violates established procedures
  (in particular the human operator)?
• When an unexpected event occurs?

8
Relationship with the human Operator

• The operator
•  Black box  intentions unknown
• Information from observation
• But from the agents point of view
• Some operators abilities are known
• Task Model
• Interaction on the agents request
• To accomplish a Task (delegation)
• Validation / Veto

9
Three Main Functions

• From the agents side
• An Adaptive Autonomy structure based on three
  core functions
• Planning
• Situation Assessment
• Authority Sharing

10
Planning

• The planning function
• Allocates and Schedules tasks between entities in
  order to reach mission objectives
• Updates parts or totality of the global plan if
  needed (replanning) in case of disruptive events

Possibility to optimize chosen criteria (e.g.
performance, operators time, etc.)
11
Situation Assessment

• The Situation Assessment function
• Constantly analyses the current state of the
  system
• Predicts the future states of the system
• Detects inconsistencies between predicted states
  and observed states (conflicts)

Example a tasks failure
12
Authority Sharing

• The Authority Sharing function
• Identifies conflicts detected by the situation
  assessment function
• Solves conflicts if necessary with the planning
  function

Several abstract solving methods available which
one to choose?
13
Basic concepts

• Mission basic concepts
• Resources (including tasks)
• Sources of resources
• Conflict

14
Resources

• An essential concept of the formalization
• Resources
• represent all items needed for the mission
  accomplishment
• are defined over time.
• Examples of resources
• Use of physical objects (sensor, fuel, etc.)
• Use of immaterial objects (pieces of information,
  logical conditions, etc.)
• Tasks

15
Resources tasks

• Tasks themselves are resources.
• A complex task can be divided into subtasks,
  which are resources.

Example of a task instanciation nav1 lt
navigating, task, map, navAlgo, waypointList,
src gt
16
Source

• A source defines the producing entity of a
  resource.
• source lt tprod , e gt
• with tprod the production time
• and e the producing entity.
• There are several possible entities
• operator
• agent
• external world
• procedures

17
Conflict

• Informal definition
• an observed or predicted inconsistency
• disturbing the predicted execution of the plan
• occuring at the ressource level.

18
Conflict
Agents plan

• Example

r1
r2
r3
R3

• Task failure

• Operators action

Task1
Failing
r4
r5
Task2
19
Conflict
20
Conflict Solving

• Initiative decision of solving the conflict
• Solving Method
• Depends on the involved entities
• Temporary or definitive
• A consistent plan
• Examples
• tasks reallocation
• interaction

21
Experimental Setup

• GPS
• Odometry
• Inertial sensors
• Ultrasounds sensors
• Scenic camera
• Procosa

22
Conclusion

• Resources to model an agents plan
• Detection and identification of conflicts at the
  resource level
• Conflict solving
• Authority Sharing ? Conflit Detection Solving
• A work in progress
• Formalization to develop (Petri nets)
• Interface design
• Experimentations