FIELD AND SPACE ROBOTICS LABORATORY
Physics-Based Design, Planning, and Control
of Robotic Systems in Space
Chris Brooks, PhD student
Drs. Paul Schenker, Terry Huntsberger, Eric Baumgartner, Samaad Hayati, and Rich Volpe, Jet Propulsion Laboratory
Future planetary exploration missions will require small mobile robots ("rovers") to travel long distances through challenging terrain, with limited human interaction. To accomplish these objectives, future control and planning methods must consider the physical characteristics of the rover and its environment, to fully utilize the roverís capabilities. Current motion planning and control algorithms are not well suited to rough-terrain, since they generally do not consider the physical capabilities of the rover and its environment. Failure to understand these capabilities could lead to unnecessarily conservative behavior, or endangerment of the rover. This research program aims to develop physics-based algorithms to allow rovers to safely traverse rough terrain with a high degree of autonomy.
Over the past several years, the FSRL has developed numerous control and motion planning algorithms designed to increase rover mobility and safety in rough terrain. Estimation of terrain physical properties is an important aspect of all of this work, since rover wheel-terrain interaction plays a critical role in rough-terrain mobility. For example, a robot traveling through loose sand has very different mobility characteristics than one moving across firm clay. Recently the FSRL has developed several algorithms for estimating terrain geometry and physical properties using on-board rover sensors (see Figures 1 and 2). These estimates are used as inputs to improve the performance of various control and motion planning algorithms.
Figure 1: Free-body diagram of rover wheel for terrain estimation algorithms.
Figure 2: FSRL terrain characterization testbed used for rover-terrain interaction studies.
A rough-terrain control method was developed that improves rover ground traction and reduces power consumption (see Figure 3). The algorithm optimizes individual wheel torque based on multiple optimization criteria, which are a function of the local terrain profile. A key element of the method is to be able to include estimates of wheel-terrain contact angles and soil characteristics. Simulation and experimental results for a micro-rover traversing challenging terrain have demonstrated the effectiveness of the algorithm.
Figure 3: Diagram of rover for traction control algorithm.
Robots with actively articulated suspensions, sometimes called "reconfigurable robots," can improve rough-terrain mobility by modifying their suspension configuration and thus repositioning their center of mass. One example of an articulated suspension robot is the Jet Propulsion Laboratory's Sample Return Rover (SRR) (see Figure 4). The SRR can actively modify its two shoulder joints to change its center of mass location and enhance rough terrain mobility. The FSRL has developed a method for optimizing the stability of such systems in rough terrain. This algorithm allows articulated suspension rovers to traverse very steep and rocky terrain (see Figure 5).
Figure 4: JPL SRR, a reconfigurable rover (courtesy JPL).
Figure 5: JPL SRR during rough-terrain traverse. Click here for movie (130 MB).
The FSRL has also developed a method for rover motion planning in rough terrain. Many "traditional" robot motion planning methods cannot be successfully applied to the rough-terrain planning problem since they ignore vehicle kinematics and dynamics, assume perfect knowledge of the environment, and represent obstacles and free space in a binary manner. We have developed a motion planning method that considers terrain, modeling, and path-following uncertainty, and also utilizes both kinematic and force analyses of rover-terrain interaction. The method is composed of two steps. The first step rapidly plans a path through the range map from the rover start position to the goal position. The second step is a rigorous evaluation of the proposed path using a physical model of the rover. The method has been evaluated in both simulation and experimentation.
All rights reserved.