Reducing Risk and Increasing Exploration Payoff with Symbiotic Rover Pairs

The planetary regions that offer the greatest scientific value are often also the riskiest to explore. The caves and canyons covering the surface of the Moon present safe havens from radiation and micrometeorites. The craters on the Lunar poles show promise for harboring volatiles such as water and hydrogen. Recurring slope linea on the steep cliffs of Mars may contain evidence of past life. All of these are prime locations for future missions, but contain a significant amount of risk during their exploration. Much of this risk stems from the current approach of sending a single rover as the sole mobile mission asset to explore these regions. The Planetary Robotics Laboratory at Carnegie Mellon University has proposed the idea that Symbiotic Exploration, multiple rovers complimenting each other's strengths and weaknesses will be far more effective at exploring these regions while significantly mitigating the risk that comes from single-rover architectures. Multiple rovers allows missions to explore more area, take greater risks, and reap better rewards. While the mechanical challenges of Symbiotic Exploration has been studied through such rovers as JPL's Axel and the Tokyo Institute of Technologies SMC Rover, the algorithmic challenges have been less thoroughly explored, specifically with regards to symbiotic path planning. These challenges include resource constrained planning for operating in light and power sparse environments, multi-agent resourcebased rendezvous for dropping off samples and recharging power, and maximum separation distance during routes to enable constant communication between rovers. This research sought to develop a multi-rover path planner capable of generating routes for multiple heterogeneous rovers that address the above constraints inherent in symbiotic exploration. To do this, an iterative multi-rover path planning algorithm known as Distributed Path Consensus (DPC) was adapted to handle symbiotic constraints. The core idea behind DPC is to first plan unconstrained paths for each rover to their goals, and then make slight adjustments to these paths through a weighted penalty function that draws the rovers closer together at desired points in time. A key challenge with DPC is that it is only able to enforce constraints based upon a known point in time. In the case of a rover expending energy, the moment in time when it will require recharging is not known. Furthermore the value at which the constraint is applied must be identical across both rovers. There is no way of having the robots rendezvous when one of them has expended a certain amount of energy, regardless of when or where that occurs. This research proposed an enhancement to the DPC algorithm (known as DPC.TF) to address these shortcomings. This was done through the definition of two classes of constraints: Feeder constraints that apply when a certain condition is met for a rover (eg. it has traveled 30 meters or has expended 1 kWh of energy); and Tanker constraints that apply when the partner's constraint is met. Through these two classes of constraints rendezvous based upon resource expenditure, as well as maintaining communication distance between two rovers can be enforced. DPC.TF allows for complex constraints between multiple rovers could be easily specified to allow rovers to plan paths to different goals while rendezvousing to replenish battery levels, drop of samples, and communicate data. This capability presents a major upgrade from the state of the art DPC algorithm which only allowed for rendezvous between rovers at predetermined times, rather than dynamically based upon research expenditure. This algorithm, and its corresponding planner, were used to show that routes on the Lunar south pole do exist that could be used to explore the permanently shadowed regions most likely to harbor volatiles.