Using MSC Software’s Adams multibody dynamics software, NASA’s Jet Propulsion Laboratory (JPL) engineers simulated the final sky crane landing sequence of the Curiosity Mars rover. The sky crane landing sequence required the rover to transform from its stowed flight configuration to landing configuration while being lowered to Mars, wheels-down, from the descent stage.
The rover, the most critical part of the simulation, was modeled to a high level of fidelity, including many flexible elements, some of which incorporating nonlinear stiffness and damping. The descent stage model is much simpler, consisting entirely of rigid bodies. In the beginning of the project, separate models were used for the rover separation, mobility deployment, and touchdown phases. During the later stages of the project, all of the models were merged into one. The combined model runs between 17 to 93 minutes on a four-CPU Hewlett Packard UNIX workstation.
Adams was used to predict the loads on components and subassemblies, and the loads in turn were used as input for structural analysis. The simulation optimized the design to provide the strength to withstand mission loads while minimizing size and weight. The philosophy of the modeling was not to try and predict every event to 100% accuracy, but to determine the bounding limit design loads that could be expected on every component.
Prior to the Adams simulation, JPL engineers believed that the projected 1-meter-per-second maximum speed during touchdown would not induce particularly high loads on the rover. Simulation, however, showed that the loads were much higher than expected. The original expectations also were that the rover would be in a quiescent state when it touched down on the Martian surface. The Adams simulation revealed that the rover was actually rotating and swinging as it landed. As a result, the rover structure was stiffened to mitigate these issues.
Later studies provided the additional surprise that the deployment of the rover’s wheels and struts, as originally planned, generated even greater loads on some of the rover components than the touchdown. Simulation showed that the end of wheel deployment generated hammer-like blows on the rover suspension and frame. JPL engineers addressed the problem by changing the timing of the wheels and strut deployment. The adjustment also reduced the swing rate and swing angle before touchdown.
The most critical aspect of the separation between the rover and descent stage was the need to avoid contact between the flight hardware. The Adams simulation checked the clearance and ensured that there was no possibility of contact. In the final design, there was very small clearance but no contact issues.
The writers of the flight control software, written in C++, required a detailed mechanical model to accurately predict the system performance. The engineers overcame the issue by compiling the controller with the Adams solver. The compiling made it possible to validate the system performance and tune the controller parameters with a detailed mechanical model. JPL engineers validated and updated the Adams model by correlating the simulation results to the test data.
Because the JPL team could not test most of the critical mission events on Earth, they had to rely upon simulation to design most of the critical hardware and control sequences for this mission. The accuracy of these simulations was proven by the success of the mission: Curiosity landed safely on Mars on August 5th, 2012.