Control of Multiple Autonomous Robot Systems

1
Control of Multiple Autonomous Robot Systems
GRASP Laboratoryhttp//www.cis.upenn.edu/mars

• Vijay Kumar
• Camillo Taylor

Aveek Das Guilherme Pereira John Spletzer
2
Multiple Autonomous Robots

• Hybrid Systems Approach to Robot Software
• modes as behaviors
• composition of modes
• Cooperative Control of Multiple Robots
• cooperative manipulation
• formation control
• tracking
• pursuit
• Human interaction
• visualization

3
Vision for Multi-Robot Teams

• Mobile platforms for deploying cameras into an
  environment
• The case for cameras
• Small
• Cheap
• Passive
• Low Power
• Uses for imagery
• Visualization of remote environments
• Obtaining information about targets
• Position, level of activity etc.
• Basis for convenient human robot interfaces

4
Visualization of Remote Environments

• Registered omnidirectional images can be used to
  visualize remote scenes

5
Visualizing the scene

• Scene can be interactively explored and/or
  revisited with a new camera trajectory specified
  by the user

6
GRASP Laboratory
7
View Synthesis with Quasi-Sparse Correspondences

• Dense correspondences can be difficult to obtain
  due to..
• Occluded regions
• Homogenous image regions
• Strategy
• Focus on accurately reproducing the motion of
  edges in the scene
• Use interpolation to estimate the motion of the
  other points
• Basis for visualization in MARS 2020

8
Novel view movies
9
Freespace Reasoning

• We can reason about the structure of space by
  considering the union of the freespace volumes
  induced by a collection of triangulated disparity
  maps.

10
Results with 3D reasoning
11
Multi-Eyed Stereo Systems

• Locations of targets and objects in the
  environment can be deduced from image
  measurements acquired by the robots
• The robot team can effectively be viewed as a
  multi eyed stereo system

12
Sensor Planning and Control

• Interesting property of these robot teams,
  estimates for various parameters of interest are
  obtained by combining measurements from multiple,
  distributed sensors
• We could choose to view our team as a multi-eyed
  stereo system where the eyes can be moved
• Question
• Given that the sensor platforms are mobile, how
  should they be deployed in order to produce the
  best estimates?

13
Theoretical Framework
CR denotes the robot configuration space, and r
is an element of CR and denotes an element of
this configuration space CW denotes the feature
configuration space, and w is an element of CW
and denotes an element of this configuration
space denotes the measurements obtained by
the robot team
r x1, y1, q1, x2, y2, q2T
w xt, ytT
xt, ytT
a1, a2T
x2, y2, q2T
y
a2
q
x
14
Optimization Problem
is a function that provides an estimate of
the feature state given the robots configuration
and the sensor measurements is a
function that returns the expected error between
the estimate returned by Est and the actual
feature state for a particular robot
configuration, r. This will depend upon our
model of sensor errors P(w) is a probability
density function on CW
Given this terminology, one can define a quality
function which reflects the
expected error in estimating the feature state
from a given robot configuration r
Objective
15
Computational Approach
The optimization problem of minimizing Q(r) is
typically difficult to solve analytically
Particle Filtering approach Approximate P(w)
by a set (wj , pj), where wj is a single sample
from CW , and pj a weight reflecting the
probability of wj representing the state w. The
integral can then be approximated by a tractable
summation.
The resulting function is typically piecewise
continuous in r and can be optimized using
standard techniques
16
Implementation Example
17
Integrating Sensing and Control

• Piggyback on particle filtering approach for
  sensing to obtain
• the particle set ?i,?i
• Framework offers a complementary relationship
  between
• sensing and control

18
Tracking Targets
19
Tracking Targets contd
20
Handling Obstacles
21
Technology Transfer and Transition

• Robot hardware and software
• University of Colorado
• Oklahoma State University
• Georgia Tech
• Evolution Robotics
• Jim Ostrowski
• DoD Programs
• MARS Teams (2020)