03/03/2013
Matt Cross
Preliminary Science Report:
Robotic Assistant for Human Explorer
This summary discusses the preliminary results of EVA observation of field geologists in simulation. The observed human explorer has extensive experience in simulation and conducting field geology while wearing a space suit. The purpose of the observation was to determine the amount of time spent during various stages of an EVA to determine the extent to which a robotic assistant would improve the overall efficiency of the human explorer. These observations will then feed into the overall concept and proposal for a robotic assistant.
The EVA 3 took place on Sol 5, and EVA 7 on Sol 8. It should be noted that EVA 7 was interrupted by the presence of the media crew, however the observations will still be valid even if the time allocation is not. EVA 3 was approximately 3 hours in length from airlock departure to airlock re-entry. An additional hour was required for EVA preparation and return procedures, such as putting on and taking off the suits, radio checks, etc, and are not considered as part of the study. Approximately 2 hours of the EVA was transit time driving the ATVs to and from the parking site, and then hiking to and from the canyon in the Summerville formation. During these 2 hours, several radio checks were made to determine where it was possible to make contact with the Hab. The remaining hour was spent collecting 3 samples and transit between sample locations. The return hike back to the ATVs was delayed due to some uncertainty in the path back.
The observed human explorer required approximately 15 minutes per sample collection after a target was identified. The process for collecting the sample went as follows: setting down equipment bag, placing the scale bar next to the sample, taking context and close-up photographs, taking a GPS measurement, writing down the GPS coordinates and time in a notebook, removing sampling tools and sample bags from the equipment bag, writing notes in the notebook, and scooping the samples and placing in the sample bags. Of the 15 minutes spent collecting samples, 5 minutes were spent scooping the samples and placing them in the sample bags.
The canyon at the Summerville formation reduced the range of the radios; the human explorers were not able to communicate with the Hab while in the canyon. It was only when out of the canyon and in line-of-sight of the Hab were the human explorers able to communicate. This did not result in any specific time delay; however the lack of communication could impede the performance of the explorers should they have needed to communicate, such as during a health and safety emergency.
EVA 7 involved two human explorers examining an inverted channel. The first explorers climbed to the upper level of the inverted channel to act as a reference scale while the second took context photographs of the first from around the perimeter of the inverted channel. This process was interrupted by the media crew on several occasions. However, it is estimated that this process would take approximately 10 minute to complete without any interruption. On two occasions the first explorer requested the assistance of the second to climb to the upper level to examine specific features in the inverted channel. The second human explorer then had to pause the context photography and climbed to the upper level, and then climbed back down. On each occasion this was an additional 5 minutes.
An important criteria in sending humans to explorer extraterrestrial surfaces will be maximizing scientific return. In other words, the goal is to maximize the amount of samples the human explorer is able to study as a function of the cost of sending the human to an extraterrestrial surface. From the above observations we can make recommendations for using robotic assistants, both mobile and static robotic agents, for improving the efficiency of human explorers:
1) Robotic agents as waypoints and communication relays. A robotic agent can be deployed at positions of known communicable line-of-sights with the Hab to act as relays when the human agents traverse further from the Hab. The robotic agent can also then act as a waypoint for the human agent to return to if they are taking that return path back to the Hab.
2) Robotic agents as guides. Mobile robotic agents can be visually taught to repeat a path back to a starting location [ref: T. Barfoot] (such as a waypoint robotic agent). This guide would allow the human agents to return to their starting point, such as the parked ATVs, without losing their known return path. A further extension of this guide would be the use of vibrotactors to provide directional cues to the human agent in the event of reduced or impaired visibility.
NOTE: one of the proposed tests at MDRS was to use a pair of vibrotactors to provide tactile cues to allow a suited human agent to follow a mobile robotic agent. The software to wirelessly connect the vibrotactors to the Android OS tablet was not ready in time for deployment. Furthermore, the Bluetooth enabled TI chips used to activate the vibrotactors are subject to export control and it may have caused problems crossing the border between Canada and the US. The vibrotactors do not specifically require the MDRS sim environment to test and it was decided that using directional cues provided visually by the Android Os tablet would suffice. This test was not conclusive as the errors from the onboard sensors were too large to provide useful directional cues. Furthermore, the use of GPS is not realistic for extraterrestrial surface exploration. Other means to determine the relative position of the human agent to the mobile robotic agent will be explored following the EuroMoonMars campaign.
3) Robotic agents as context imager and localizer. The human agent, while collecting samples or scene mapping with another human agent, took taking images. The robotic agent would take the context photographs and localize the positions from where the images were taken and the positions of the targets. In the case of EVa 7, the mobile robotic agent would travel around the base of the inverted channel while both human agents explored the upper level. Should the human agent want images of a specific spot, they can use gestures and communicate with the robotic agent.
NOTE: the gesture capturing study will be completed between Sol 10 and Sol 12. The data will be collected for post-processing after the EuroMoonMars campaign however qualitative results will be provided following the tests.
4) Robotic agent as field assistant. The human agent takes notes relating to the location of the sample. These notes include a description of the location, the time, the GPS coordinates, and all other notes related to collecting the sample. While this would not be specifically a robotic tool, having voice recognized note taking would allow the human agent to dictate notes while collecting the samples in parallel instead of in series. In addition, the robotic agent would add the localization, times, and context images to the notes. With the notes with images and localized data already captured, the human agent would be able to perform light editing to complete reports of the samples instead of having to prepare an entire report.
Using EVA 3 as a baseline, the following improvements could be imagined:
1) Less travel time reaching the canyon as fewer radio checks would be needed with radio relays.
2) Sample collection would reduce from 15 minutes to 5 minutes per sample with robotic context imagers, localizers and note taking. Thus, in the hour spent in the canyon additional samples could be collected instead of planning an additional EVA of 3+ hours to collect additional samples.
3) The return travel back to the ATVs would be reduced as the known path would be followed. This would allow the human agents to return to the Hab sooner; with less time required for report writing and more time available at the Hab, more time can be devoted to sample analysis.
Point 2 is the most important as EVAs can be saved for different sites.
A final report will follow the motion capturing.