1. Multiple Robots For Multiple Missions: Architecture for Complex Collaboration
Noa Agmon1, Oleg Maksimov1, Ariel Rosenfeld2, Shai Shlomai1, Sarit Kraus1
1Bar-Ilan University, Israel 2Weizmann Institute of Science, Israel
AGA IJCAI 2017

2. What do you want from us?
Multi-robot systems require flexibility.
New missions
Form new teams
React to environment changes and failures (e.g., communication)
Human supervisor
Etc.
No existing architecture is able to handle all of the above in real deployment.
We propose a 6-layer architecture and empirically evaluate it.
In multi-robot systems such as search and rescue or warehouse operation flexability is essential.
Namely, the robots should be able to efficiently accommodate new missions, new team formation, react to environment changes and faluts – for exa=ple lack of communication and so on.
To date, no exsisting arch. Is able to handle all of the above challenges in real deployment.
Therefore, in our paper we propose a 6-layer arch.. And implement it to exemplify its benefits.
AGA IJCAI 2017

3. A robot may be assigned to more than a single team.
Our arch is of 6 layers.
At the bottom we have semi-autonomus robots, which can be grouped to team.
A TEAM is a virtual structure that allows members to share information and synchronize easily.
A robot may be assigned to more than a single team.
One or more teams are assigned with a mission. For example, peforming a patrol or a serach over a terrain. The mission manager is thus inchare of sync. The teams and monitoring the resources, information etc.
All missions of the sys are managed by a system manager which moniters all the missions and controls the data flow between them.
A supporting agent can also be used to mitigate the information to the UI such that the information is prioritized and the human operators that use the UI maximize their capabilities.
We disscuss these layers next.
AGA IJCAI 2017

4. Layer 1 - Robots Each robot has “built-in” functionalities.
Rotate wheels 20 degrees left
Navigate to GPS coordinates (x,y,z).
Etc.
Each robot is semi-autonomous.
At the bottom…
AGA IJCAI 2017

5. Layer 2 – Team Manager
Robots may be clustered into teams which in turn can be clustered into teams which….
The team manager manages a group.
Allows inter-team communication
Represents the group’s state
Etc.
AGA IJCAI 2017

6. Layer 3 – Mission Manager
Oversees a mission.
Creates PLANS for the subordinate teams.
The Plans are overseen by the
team manager(s) which report to the MM.
AGA IJCAI 2017

7. Layer 4 – System Manager Monitors and creates all
Missions
Resources
Etc.
Usually, centralized?.
AGA IJCAI 2017

8. Layer 5 –Supporting agent
Information filtering and prioritization
for the human operator.
Smart buffer between the system
manager and the human operator(s)
MOVIE!
Rosenfeld et al. Intelligent Agent Supporting Human-Multi-Robot Team Collaboration, IJCAI 2015 (AIJ 2017)
AGA IJCAI 2017

9. Layer 6 – UI
AGA IJCAI 2017

10. Evaluation PLACE ROBOTS SCENARIO
Outdoor environment (two basketball courts). A total 25x25 meters.
ROBOTS
The system consisted of four Hamster robots. Each is semi-autonomous robot, that cooperates according to the architecture described below.
SCENARIO
The operator decides to execute a patrolling task, and defines the waypoints for the mission.
Then, communication problems occur.
AGA IJCAI 2017

11. MOVIE
AGA IJCAI 2017

12. Results Idleness = time between two visits to a patrol point.
Sync = overhead
AGA IJCAI 2017

13. Results 2 - Simulation Good scalability in environment size
Increasing the size of the patrol by 2.5, 5 and 10. Examined 15 and 20 robots patrolling.
Good scalability in environment size
AGA IJCAI 2017

14. Summary New Architecture. Allows ample flexibility and resilience.
Empirical evaluation support that.
All code is available online:
AGA IJCAI 2017

15. All code is available online:
Thanks
Ariel Rosenfeld
All code is available online:
Noa Agmon, Oleg Maximov, Ariel Rosenfeld, Shai Shlomai, Sarit Kraus
Multiple Robots For Multiple Missions: Architecture for Complex Collaboration
AGA IJCAI 2017