|
Post by Hannes Vilhjalmsson on Mar 2, 2011 20:20:21 GMT -5
This week we are reading an article written by our own rector Ari Jónsson and his colleagues at NASA before he joined Reykjavik University. Here is the reference: Activity planning for the mars exploration rovers by John L. Bresina , Ari K. Jónsson , Paul H. Morris and Kanna Rajan. In the proceedings of the 15th International Conference on Automated Planning and Scheduling, 2005. This is part of the incredibly exciting and historic NASA Mars Rover project. Please provide your questions (or dicussion points) for Ari below.
|
|
|
Post by carmine on Mar 5, 2011 10:24:24 GMT -5
- in the Introduction there is a section about "key characteristics" (optimization, temporal constraints..); they affect the goal of the search. For each of them, is there a specific weight (defining a priority) or all of these characteristics have the same importance? - can you give us more details about the "dynamic backtracking" technique (an example can clarify the concept)?
|
|
|
Post by lorenzo on Mar 5, 2011 14:53:34 GMT -5
- The mapgen is still improving/ updating with new feature (like new algorithm, new component etc...) ? If yes which ones?
- This system is used only for the project MER or it is used also in other project that request this kind of framework? If yes which one?
|
|
|
Post by finnur on Mar 6, 2011 7:11:51 GMT -5
When the MER first arrived on mars, how did it start plan when it knew nothing about it's environment, or did it perhaps have some knowledge of it beforehand?
|
|
|
Post by Ásgeir Jónasson on Mar 6, 2011 8:21:00 GMT -5
1.
What information about the environment is taken into account when a plan is constructed ? Is the footage from the rover only utilized or are pictures taken from satellites of nearby surface also taken into account ? And also regarding the evaluation of the footage, does the rover have a database of geological facts or is the footage evaluated on the ground ?
2.
The article does mention a lot of reasons for involving humans in the rover's planning, like the fact that assessing a plan constructed entirely by it is very difficult. Do you think that you will be able to omit the human factor of the planning sometime in the future or are the problems presented in the paper impossible to overcome or is it maybe essential to have a human overseeing the plan making, regardless of future advancements ?
3. Have any significant changes been made to the MAPGEN system since this article was written ?
|
|
|
Post by thorsteinnth on Mar 6, 2011 12:23:26 GMT -5
1. Það er sagt að roverinn treysti ekki bara á gervigreind heldur fái hann skipanir frá jörðu um hvað eigi að gera. Hversu miklar hömlur verður þá ljóshraðinn og tíminn sem það tekur fyrir skipanir að fara fram og til baka?
2. Það var tekið fram (undir Heuristic search guidance) að leitin myndi velja verk með hærri forgang fram yfir verk með lægri forgang. Hvað myndi það þýða ef eitthvað verk með háan forgang þyrfti að keyra annað með lægri forgang til að geta klárast?
|
|
|
Post by kristjan on Mar 6, 2011 15:43:56 GMT -5
It is mentioned that the rovers are landed on widely separated sites. Is it then one rover per site or are there more rovers than that?
How expensive is it to build a rover compared to the total cost of the project of sending rovers to mars and having them obtain scientific data?
How much more difficult would it be to maintain more rovers than where sent for the project?
|
|
|
Post by Jón Þór Kristinsson on Mar 6, 2011 16:33:27 GMT -5
Has the MAPGEN system been used to plan research in labs such as the one on the International Space Station? or even in more simpler ways such as project planning (construction, programing and such)?
|
|
|
Post by grimurtomasson on Mar 6, 2011 16:47:15 GMT -5
To what other problem domains may mixed-initiative planing be more appropriate than completely automated planing? What do they have in common?
Have similar mixed-initiative planning systems been devised outside of space exploration?
What key lessons did you take away from this project?
|
|
|
Post by Hrafn J. Geirsson on Mar 6, 2011 16:48:05 GMT -5
1) How does the operations staff go about generating the daily activity plans ? In general which kinds of decisions are made by humans, and which are made by the AI ?
2) "the search was biased to work on high priority activities before low priority ones. Virtually no other heuristics were used for determining the order in which open decisions were tackled."
Were there only two priority classes for activities (high and low) ? Surely, some of the activities must be more risky, time consuming or resource consuming than others. How come no other heuristics were used ?
|
|
|
Post by Elín Carstens on Mar 6, 2011 17:00:59 GMT -5
It is mentioned that an attempt to automatically resolve resource violations only had limited success, in part due to the difficulty of obtaining sufficient information about resource usage from the resource computation modules.
1) What would you have considered a total success?
2) What kind of models did you use for resource management?
|
|
|
Post by Eiríkur Fannar Torfason on Mar 6, 2011 17:08:29 GMT -5
1. Can you tell us a bit more about the algorithms involved? The article mentions using Bellman-Ford for checking for Arc Consistency but I thought something like AC-3 would be used for that. Also, the dynamic backtracking search sounds interesting.
2. The article suggests that many of the design decisions were geared towards providing the users with a sense of control, rather than perhaps find an optimal plan. How actively did you engage the users while developing and testing the system?
|
|
|
Post by Þorgeir Karlsson on Mar 6, 2011 17:13:05 GMT -5
This is a bit off topic but in the paper it is mentioned that the rovers have UHF antennas that communicate with satellites orbiting Mars. I wasn't aware there were satellites orbiting Mars. Could they be used to create something similar to Google Earth? Have NASA considered expanding the rover missions by sending rovers to other planets or moons beyond Mars, or would the fact that they are further away create too many problems?
|
|
|
Post by Jökull Jóhannsson on Mar 6, 2011 17:25:44 GMT -5
1. How much ground do you estimate the rovers have explored ? 2. What is the biggest discoviry made by these rovers
|
|
|
Post by questions on Mar 6, 2011 18:24:07 GMT -5
við það að lesa textan þá rakst ég á tvö hugtök sem ég skildi ekki getur þú vinsamlegast útskýrt fyrir mér hvað þessi hugtök standa fyrir
1 .quantitative reasoning 2. Cartesian Product
einnig er ég ekki alveg klár á því afhverju þeir nota bæði APEN og EUROPE hver er ástæðan fyrir því
bonus spurning eftir að lesa þessa grein þá er ég viss um að það sé náungi hjá NASA sem gerir ekki neitt annað en koma up með skamstafanir fyrir forrit og plön hvað er starfsheitið hjá þeim einstaklingi
|
|