Title: Improving the Performance of Mobile Robots on Uneven Terrain
1Improving the Performance of Mobile Robots on
Uneven Terrain
- Joseph Auchter
- Dr. Carl Moore
- FAMU/FSU College of Engineering
- 9 October 2007
2Wheeled Mobile Robots
- Increasing interest in autonomous robots
operating outdoors on difficult terrains
GDRS XUV
NASA Spirit Rover
CMU McArthur
GDRS TAC-C
3Wheel Slip for Outdoor Robots
- All wheeled vehicles will slip on uneven terrain
- Two kinds of wheel slip
- Dynamic due to insufficient friction, terrain
deformation, etc - Kinematic due to lack of an instantaneous center
of rotation compatible with all wheels
4Why Slip Occurs on Uneven Terrain
- Example Ideal Ackermann Steering
Wheel / Ground Contact Points
5Why Slip Occurs on Uneven Terrain
- Ackermann Steering on Uneven Terrain
Wheel / Ground contact point locations vary
because of uneven ground
6Problems Caused by Wheel Slip
- Decreased localization ability due to odometric
(wheel encoder) error accumulating without bound - Power wastage
- Reduced traction, terrain traversibility
7Wheel Slip Example
8Effects of Wheel Slip Example
- Huntsberger, et al (2002) long-range rover
autonomy - Application to NASAs Spirit and Opportunity Mars
rovers - Found 15 error in odometry over test run
- Reported wheel slip resulting in increased power
consumption
9The Proposed Solution
- Concept by N. Chakraborty and Dr. A. Ghosal at
the Indian Institute of Science (2003) - Passive Variable Camber (PVC) Lateral tilting of
wheels allows the robot to move on uneven terrain
without kinematic slip
10Research Hypotheses
- PVC will significantly reduce kinematic slipping
on uneven terrain - PVC will allow the wheel or tire to maintain
better contact with the ground, improving
traction and reducing dynamic slip - PVC will reduce power consumption
11Kinematic Simulation of a WMR
- Uneven terrain
- Robot equipped with Passive Variable Camber (PVC)
joints - Traditional robot modeling is inadequate
- Need a new, precise way to simulate wheels
rolling over uneven ground
12Analogy Between WMRs and Robot Hands
13Simulation Concept
- Apply dextrous manipulator modeling techniques to
a wheeled mobile robot system - Allows us to precisely simulate the motion of the
wheels on an uneven terrain
143-Wheeled Mobile Robot Model
- Front wheel is steered
- Rear two wheels have Passive Variable Camber
(PVC) joints - Robot moves on uneven terrain
15System Model
- The following ODEs describe the system
Rolling contact non-holonomic
constraints
Differentiated closure constraints
Input velocities consistent with constraints
16Robot Joint Velocities
Robot Joint Velocities
17Surface Parameterizations
18Contact Variables
- Contact variables for one wheel
- Grouped for all wheels
- Velocities of the wheel relative to the ground
19Rolling Contact Equations
- and Vc are related by Montanas1 equations of
contact
1 Montana, D. 1988. The Kinematics of Contact
and Grasp. The International Journal of Robotics
Research, Vol.7, No. 3, 17-32.
20Rolling Contact Equations
Contact point on wheel
Contact point on ground
- CK relates the wheel velocity to the motion of
the contact point on the wheel and ground
surfaces.
1 Montana, D. 1988. The Kinematics of Contact
and Grasp. The International Journal of Robotics
Research, Vol.7, No. 3, 17-32.
21Closure Constraints
- Robot / ground system is a hybrid series /
parallel mechanism. - There are three serial kinematic chains in
parallel - Closure constraints each chain of coordinate
transformations must end in the same frame (P
in this case)
Chain 1
Chain 2
Chain 3
Purple intermediate frame
22Closure Constraints
- The closure constraints can be written in the
form
- Group the configuration variables together as
- Differentiate the constraints to make ODEs
23Velocity Relationships
Platform velocities
Wheel velocities
- Constraint equation originally developed for
modeling dextrous robotic manipulators
2 Han, L., Trinkle, J.C., and Li, Z.X. 1997.
The Instantaneous Kinematics and Planning of
Dextrous Manipulation. Proc. 1997 IEEE Intl.
Symp. on Assembly and Task Planning, pp. 60-65.
24Velocity Relationships
Platform and wheel velocities
Joint velocities
- Want to choose input velocities which are
consistent with these constraints.
25Velocity Relationships
- Start with the velocity constraints
- After some manipulation, we can write
Desired input velocities
Input velocities consistent with the constraints
26System Model
- The following ODEs describe the system
Rolling contact non-holonomic
constraints
Differentiated closure constraints
Input velocities consistent with constraints
27Kinematic Simulation Results
28Kinematic Simulation Results
29Simulation Results (Hill Climbing)
30Simulation Results (Hill Climbing)
31Simulation Results (Random Terrain)
32Simulation Results (Random Terrain)
33Simulation Results (Random Terrain)
34Next Steps
- Path planning for the robotic system using the
kinematic model - Design and construct experimental test-bed
- Show that a wheel with Passive Variable Camber
can roll over an uneven terrain without kinematic
slip - Investigate effects of PVC on power consumption
and dynamic slip