Principal Investigator Steven Dubowsky
Co-investigator Karl Iagnemma
Project Website http://robots.mit.edu/projects/mars/index.html
Mobile rovers will be a primary platform for studying the surface of Mars during the next several years. Substantial research effort has gone into developing autonomous navigation and path planning algorithms. Experimental analysis of these algorithms has shown that understanding rover wheel-terrain interaction is very important for accurately controlling rover motion and for preventing loss of mobility.
The goal of the research is to develop on-line algorithms to characterize terrain and predict terrain traversability. We have developed a parameter estimation method for characterizing the terrain. As an input to this terrain characterization algorithm, a method for visually measuring wheel sinkage into soil was developed. We are currently working on several other methods for terrain characterization and classification.
Parameter-Based Terrain Characterization -- The traversability of a certain type of terrain for a wheeled vehicle can be characterized by 2 parameters, ground cohesion c and internal friction angle f. We have developed a method for estimating these parameters online, using on-board rover sensors. A soil-tire interaction test-bed has been developed for algorithm validation.
Visual Wheel Sinkage Measurement -- Wheel sinkage into the soil is an important input to the above terrain characterization algorithm and is also an important indicator of vehicle mobility. To this end research has been conducted toward finding an efficient method for measuring the sinkage of a wheel into deformable terrain.
An algorithm was created which uses configuration sensors and an image from a rover-mounted camera to identify the sinkage angles. It uses configuration sensors to locate the wheel rim in the field of view, then searches along the wheel rim for the change in contrast indicative of the wheel-terrain interface.
The terrain testbed was fitted with a camera and a spare wheel from the FIDO rover. Images collected from this testbed, under various lighting conditions, were used for validating the algorithm.