At the ACFR there are regular presentations from visitors, academics and students. The first I saw was from Swiss engineer Franziska Ullrich, who presented work she had done to design an optimal Mars Rover that could take on the hostile, rocky, alien environment of an exhibit at the Powerhouse Museum.
Her approach was to drive a simulated model vehicle within a specified virtual environment (without hills above a certain pitch angle or rocks above a certain size), and measure how it performed. She changed the design and ran the simulation again to see if that design improved the vehicle’s performance: making the wheels bigger, changing the suspension, and so on. The design needs to distribute the weight of the load evenly across the wheels, making sure the wheels keep traction.
Rather than keep changing the design herself, though, Ullrich uses a genetic algorithm to try out many different permutations across the space – that is, the abstract mathematical space that maps out all possible combinations in the design. Using an approach ‘based on Darwin’s theory of evolution’, the simulator introduces random mutations for each generation. After each simulation the system selects the ‘fittest’ designs, according to the optimal design features, and abandons the designs that proved less fit. Over 100 generations, taking 3-4 hours of computing time, she improved the design by 28%.
This quite complex technical process of rocker-bogie optimisation also invokes some powerful cultural resources of storytelling and metaphor. The process of optimisation presents a story of improving the design towards some kind of perfection, with an accelerated natural selection process as the metaphorical plot path. It draws on Darwinian narratives, with their connotations of scientific legitimacy to the approach (a technique exploited most extensively by the Santa Fe Institute).
But this is a post-human design process in which the unmanned vehicle is impelled to rove alone through alternating vacuums of simulation: Mars and museum space. In which ever of these environments, it must be autonomous. The lonely rover must be able to right itself if it falls over. It must be the fittest to survive.
Leave a Reply