The year is 2050.

NASA is (finally) in the last stages of planning its first manned lunar base. They have a rough idea of where to Setup the base, but need dense surface maps to determine the best suitable location for the base, and to give its astronauts a good idea of their immediate sur- roundings when they arrive. To do this as quickly and thoroughly as possible, NASA has deployed a swarm of dozens of rovers to map the 25 km2 potential landing zone. Consider- ing the sheer complexity of having to directly monitor a large fleet of rovers, NASA has paid for Roverside Assistance to reduce the amount of required remote-human intervention with its rovers.

In addition to autonomous rover-rover interaction behaviors specifically for navigating in rough terrain which often lead to extreme dynamics, Roverside Assistance also includes inte- grated system capabilities to asses operating state of each individual rover in the fleet. Further, in case of distress situation the Roverside assistance system can autonomously queue a rescue mission to liberate an entrapped rover using a second rover as a rescuer. Finally, a pair of robots when operating as one cohesive unit generate 3D footage that is an accurate representation of each of the rover’s in-situ interaction with the environment.

 

Figure 1: NASA 2050 use case, with an image highlighting the location of the Shakleton Crater (left) and sample photo of a possible division of Shackleton Crater into segments to be searched by individual rover teams. Rover starting positions are indicated by squares (right).

 

The rovers are landed in the search zone in pairs, the plan being for each pair to map out a few square kilometers in the course of a year. One such pair of rovers, named Block and Tackle, have been assigned a search zone that includes several areas with difficult terrain that is prone to high-centering rovers who cross it. Over the course of a year, Block and Tackle follow a rough route given to them by NASA, autonomously avoiding obstacles like boulders and crevasses while building up an accurate map of their environment by cross-referencing their surroundings with available satellite imagery and their own and relative odometry. While they typically roam out of sight of each other for greater coverage, they periodically meet backup to share map data and use each other’s location to calibrate their own positions.

One day Tackle receives an SOS from Block, which reports it has just entered a boulder field and gotten high-centered. The system of rovers initiate the coordinate and rescue mode. With an updated map provided by satellite imagery and the 3D terrain map last constructed by the pair’s efforts, Tackle can autonomously route to Block’s location. As Tackle approaches the scene, the pair use Block’s odometry and information from Tackle’s sensing suite to accurately estimate Block’s relative pose and decide the best plan of action for freeing Block.

 

Figure 2: A storyboard of Block and Tackle demonstrating our system’s in action, with 4 phases: Block exploring its environment (phase 1), Block signaling for help (phase 2), Tackle successfully docked for towing (phase 3), and Block has been dislodged (phase 4).

 

The rovers determine that the best plan of action is to pull Block off of the rock against which it’s high-centered. Tackle approaches Block along roughly the same path Block used. When it is close, they dock with each other using specialized integrated booms. Once a secure connection has been established, the rovers coordinate and simultaneously move in a prede- termined direction to free Block. Within a few seconds, success! Block has been moved off of the rock it was stuck on. As a final step the rovers analyze and update their terrain maps to show that the area they are next to is unsafe for travel, and route around it as they continue on their mission!