Mars Rover Simulation with EDL
Group Project
Project Mentor: Dr. Douglas Allaire, TAMU
Fall 2014
I lead a group project during my study abroad semester at Texas A&M's main campus in College Station, TX (Fall 2014). The course was taught by Dr. 'Doug' Allaire (Past: MIT's Aerospace Computational Design Lab). The goal of this semester-long project was to simulate a Mars rover along with an Entry, Descent and Landing (EDL) system.
At the end of the semester, all the simulations were pitted against each other to see whose rover completed the landing and navigation of a randomly generated terrain quickest. Ours came Third! (Out of about 35 teams. Of course, the professor's own robot got the First place...)
Item 3a: Representative Image only. Taken from NASA.gov
Phase I: Marvin
Having nicknamed the rover 'Marvin', after the Looney Toons Martian character, we set about the preliminary system modelling of the DC motor, a simple speed reducer (gear box) and wheels, after taking into account the gravitational forces on Mars and rolling resistance. We also supplied a random terrain (as an array of terrain gradients) and studied the maximum speed generated.
Next, we simulated Marvin's power and efficiency, assuming 6 individual drive motors for each of the 6 wheels, much like a simplified version of the Curiosity Rover (which included 4 additional motors for steering). The set of functions we created would take in the mass of the body as input, and having known the mass of the motors and drive train beforehand, the total power consumed for navigating the given terrain would be output, indicating the size of the battery required.
Phase II: Power
Phase III: EDL
The centerpiece of this phase was simulating the entry, descent and landing system with focus placed on calculating the time taken from the point of entering the Martian atmosphere to the point of touch down. Obviously, the mass of the rover, including the body, engine and batteries would affect the landing time. We also created detailed documentation at this stage, for the various functions and data structures.
Phase IV: Optimization
This course being about Numerical Methods in Engineering Analysis, the last phase of the project focused on optimizing the complete system so that, given a random terrain (in the final competition), the rover would be able to land and navigate over the terrain in the least possible time, given a ceiling for the total cost of all parts used in the virtual rover. This was not a simple optimization problem - it had 14 interdependent variables, including the material used for the body (affecting mass and consequently the EDL system), the DC motor used (affecting torque and power requirements), batteries selected and many others, resulting in a 14-space of possibilities! In-fact, there was no easy way to even make sure that there exists a global optimum in that 14-space.
After a few grueling hours of trial and error, some implementation of Golden section search on crucial variables and some cunning reverse engineering ("hmm, looks like the gear box doesn't affect the power output that much, so just pick the one with the least mass"... etc.), our team came third out of 35 groups! [For a comparison, the professor's simulation of the robot that came first completed all the tasks in about 430 seconds, while ours did it in about 610 seconds].
If you're a current student in the MEEN 357 course at Texas A&M, I cannot give you our MATLAB codes (Sorry!) But, I'd love to help you with your own project!