UMass Amherst People Finder

Mike SanSoucie





        Numerous manned chemical propulsion vehicles have been designed in the past; however, there is considerably less experience for nuclear electric propulsion (NEP) vehicles.  In addition to traditional aerospace expertise, competent NEP design requires knowledge of nuclear reactors and electric thrusters as well as conversion, management, and distribution of thermal and electrical energy.  An example of a NEP vehicle is shown below.  A thorough search of the design possibilities for these vehicles is needed for a proper assessment of mission viability and identification of feasible low-mass candidates for future designs.  The Nuclear Electric Vehicle Optimization Toolset (NEVOT) was created to find candidate nuclear electric vehicle designs for a manned mission to the asteroid belt.  NEVOT is a joint effort of Marshall Space Flight Center, Oak Ridge National Laboratories, and the Arnold Engineering Development Center.  It consists of seven analysis modules: trajectory, space reactor power system, power management and distribution, electric propulsion, habitat, truss, and configuration.  A flowchart of how NEVOT is run is shown below.


        The optimization technique used in this research is a genetic programming approach.  Genetic algorithms have demonstrated an ability to find global optima over large search spaces such as this.  A genetic algorithm (GA) is a random, guided search technique that covers the global region of the design space.  As the name implies, genetic algorithms use methods analogous to biology and genetics to find optimal solutions. 


        The GA generates an initial population of designs (random or selected by the user) and defines a fitness value for each.  The fitness value is defined by means of an objective function that the user supplies based on the problem of interest, in this case minimizing mass subject to constraints.  When a population has been evaluated, the GA selects the solutions with the best fitness to undergo a genetic operation.   The strings of these designs are replicated, crossed with one another, and/or mutated to create the next generation.  Crossover occurs by swapping multiple portions of two parent designs to create offspring designs.  The crossover points and mutated variables are determined randomly, and the rate at which they occur can be defined by the user.  When the new population is filled, the process is repeated.  Thus, the genetic algorithm uses a random, directed search to find global optima.  The genetic operations allow for exploration of a large design space during simultaneous optimization.


        Mike has helped to update and refine NEVOT. He also helped add the Mars trajectories and updates to include variable specific impulse. The next step will be in creating a new optimization toolset for the NASA Crew Exploration Vehicle (CEV), and also finding funding to optimize a surface habitat.