Allocating Tasks 1
The first study addressed the planning of one day of science tasks that would be part of a 60-day mission. The agents were defined as 2 astronauts and an unpressurized rover (UPR), i.e., a robotic chassis that is teleoperated from Earth. A set of resources that would be available to these agents was defined.
The mission's activities comprised 6 science tasks, identified by NASA's Lunar Architecture Team (LAT), to be performed at each of 2 science zones on the lunar surface: collect rock samples, conduct geological context survey, collect rake sample,
collect soil sample, collect drive-tube sample, drill and collect a core sample.
The activities were divided into their component actions, each of which was described in detail to enable the algorithm to use the information.
It was assumed that each task could be performed in any of three ways:
- by astronauts working EVA (extravehicular activity – i.e., walking on the lunar surface in pressurized space suits)
- by astronauts working IVA (intravehicular activity – i.e., operating robotic arms and tools from inside the SPR's pressurized cabin)
- by a robot (UPR) teleoperated from Earth
Parameters and constraints
A table of parameters was compiled, detailing how much time would be required for each activity under each set of circumstances. Other parameters such as likelihood of success, quality of samples collected, etc., were assumed to be equal regardless of whether they were done by EVA, IVA, or teleoperation from Earth.
For example, for purposes of this study, drilling a core sample was said to take 1.75 hours for an astronaut working EVA, 3.5 hours for an astronaut working IVA, and 7.0 hours for the UPR teleoperated from Earth. (EVA times were taken from Apollo data. IVA and Earth-teleoperation times were estimated based on the EVA times, but none of these parameters has yet been validated by independent peer review.)
Constraints were identified which, among other things, limit the amount of time available to the agents. It was determined that each astronaut can perform a maximum of 8 hours of EVA activities per day. Additional time may be spent on IVA activities, but the total of EVA and IVA is not to exceed 16 hours. The pressurized vehicle and the UPR can each perform up to 16 hours before needing to be recharged at the habitat.
For EVA and IVA, it was assumed that the astronauts would be able to set up the core-sampling equipment and leave it to run while they performed other tasks. The UPR would also be able to conduct core sampling, but would have to stay with that task exclusively until it was completed.
Problem and search
The problem posed was to determine which tasks should be performed by which of these means (astronaut EVA, astronaut IVA, or teleoperated robot) and the optimal sequence of events. At the time of this first study, our sponsor gave highest priority to astronaut safety and directed us to make the objective function minimization of EVA time, since being outside in a space suit is considered to be the most hazardous state for an astronaut. The secondary objective was to minimize IVA time, and in third place was minimizing the amount of time the robot spends working.
We developed a table of starting and goal states and established a heuristic to guide the planning algorithm: As it explored all possible nodes in the search space, the planner was to penalize those which required EVAs and, to a lesser extent, IVAs. A partial graph of the search is shown below.
Partial graph of the search. "A12" means "astronauts 1 and 2." "MC2" (Mobile Chassis 2) was the term used for the UPR during the first study. "SZ" means "science zone."
We ran the planner with two scenarios: one in which the astronauts do all of the tasks themselves, and one in which the UPR does some of the work. For each scenario, HURON generated an Excel spreadsheet showing a timeline of the optimal activity plan, with actions for each entity placed on the timeline.
Since the objective function we were given was to minimize EVA time, the planner calculated that the optimal plan was simply to have the astronauts conduct all of their tasks as IVAs, without ever donning space suits and leaving the relative safety and comfort of a pressurized cabin. While avoidance of all EVA activity is not what was ultimately desired, this common-sense result validated the planner, which had to search a fairly extensive trade space to find its solution.
In the first scenario, as previously mentioned, the astronauts performed all of the tasks. In the second scenario, they performed all 6 tasks at the first science zone while the UPR performed 3 tasks in the second science zone, after which the robot needed to return to the habitat to recharge its batteries. Upon completing their work at the first science zone, therefore, the astronauts drove to the second science zone and conducted the three tasks that the UPR had left undone. The tasks that the UPR completed saved 1.5 hours of astronaut IVA time when compared to the scenario in which the astronauts did all of the work.
For more information, contact Charles Weisbin at Charles.R.Weisbin@jpl.nasa.gov