Question1:

What is an Agent?

The term "agent" has been used by many people to mean many different things. Even within the Agent Research Community, there are at least the following variants on the term agent: Mobile Agents (e.g. Voyager, Grasshopper), LearningAgents,Autonomous Agents (e.g. robots),Planning Agents, Simulation agents, Distributed Agents. In the context of the Open Agent ArchitectureTM (OAA?/SUP>), we are focused on building distributed communities of agents, where agent is defined as any software process that meets the conventions of the OAA society. An agent satifies this requirement by registering the services it can provide in an acceptable form, by being able to speak the Interagent Communication Language (ICL), and by sharing functionality common to all OAA agents, such as the ability to install triggers,manage data in certain ways, etc. In our community of agents, we are able to include, and make use of, each of the different types of agents mentioned above.

蕃薯藤搜尋網

Question2:

Honda's Humanoid Robot "ASIMO" Rings Opening Bell at the NYSE
Tokyo, February 15, 2002 --- Honda Motor Co., Ltd. (HMC) celebrated the 25th anniversary of its listing on the New York Stock Exchange (NYSE) by showcasing its technological strength with one of the most unique opening bell ceremonies in NYSE history. Honda's ASIMO, the world's most advanced humanoid robot, rang the opening bell for trading on February 14.
Honda's basic research and development in humanoid robotics began in 1986 with the goal of developing an autonomous walking robot that can be helpful to humans as well as of practical use in society. In 1996, Honda's prototype robot P2 made its debut, followed by the more advanced P3 in 1997. ASIMO, which debuted November 2000, is a further evolved version of P3 in an endearing people-friendly size, 120cm-tall, that enables it to actually perform tasks within the realm of a human living environment
The latest version of ASIMO walks more smoothly, more flexibly, and more naturally, and is able to move more freely in ordinary environments, including climbing and descending stairways and slopes. Further, it can receive voice input. Specialized movements, guidance and explanation functions now enable ASIMO to be customized according to a wide range of user requirements. This year ASIMO robots are being leased to several corporations and museums for use in greeting people at offices and events. This month the humanoid robot also started working as a receptionist for visitors at Honda's Aoyama Headquarters.
Honda, which trades on NYSE as American Depository Receipts (ADRs), opened for trading on NYSE February 11, 1977 as the first listed Japanese automaker. Honda is currently listed on nine stock exchanges around the world, including five Japanese stock exchanges (Tokyo, Osaka, Nagoya, Sapporo and Fukuoka) and four overseas stock exchanges -- NYSE, London, Euronext Paris, and Swiss SE.

詳情見

◎
◎
◎
◎
◎
◎
◎
◎

Question3:

八皇后問題解法:

Question4:

TSP

經濟時間序列

------

TSP能輕鬆地迎刃解決經濟時間序列問題，它能從任何格式包括binary、Lotus或Excel檔自由轉載原始資料，可任意轉換含有代數之時間序列，自動儲存原始及已轉換的時間序列;陳列完整表單，可於DOS、Windows及Mac、OS上繪圖，讓您隨心所欲寫成任何格式檔案。TSP使用多樣技巧計算經濟時間序列問題：最小平方法OLSQ Ordinary least squares、二階最小平方法2SLS Two stage least squares、限制最大相似法LIML Limited information maximum likelihood、 一階序列迴歸法AR1 Regression with first-order serial correlation、多項式分配延遲法

PDL (polynomial distributed lag) variables、最小絕對值導數法LAD Least Absolute Deviations、最小中間平方法LMS Least Median of Squares。

Question5:

廣度優先搜尋法 (Breadth-first search , BFS)

廣度優先搜尋法是一個簡單的搜尋策略，其根節點先被展開， 然後再展開被根節點展開的所有節點，接著再展開它們的後繼者，就這樣一層一層地展開所有的節點，直到找到目標解。簡單的說， 所有在深度(depth)為d的節點皆會比所有在深度為d+1的節點早被展開。實作時可呼叫GENERAL-SEARCH演算法，把佇列函數設為"把新展開的狀態放在佇列的後面" BFS 保證一定能找到解，如何解存在的話。(完全性)

BFS 找到的第一個解一定是路徑最短的。(最佳解)

BFS 所需的搜尋時間為：

設b為分枝因數，d為目標節點所在的深度，那最多會被展開的節點數為 1 + b + b^2 + b^3 + b^3 + .... + b^d ，所以搜尋的時 間複雜度為 O( b^d )。

它所需的時間及空間高的嚇人，有時並不適合使用，有必要再加以改良。

深度優先搜尋法 (Depth-first search, DFS) 

深度優先搜尋法總是先展開一棵搜尋樹中位於最深一層的節點，只有當搜尋遇到終節點(dead end)(它不是目標節點又沒有後繼者可展開)時，搜尋才會回溯到上一層(即其父節點)並展開較深一層的節點。實作時可呼叫GENERAL-SEARCH演算法，把佇列函數設為"把新展開的狀態放在佇列的前面" DFS 需要的空間是適度的，不會太多。如上圖所示，它只要 儲存從根節點到目前搜尋到的葉節點的單一路徑，以及在這條路徑上每一節點剩下還沒展開的兄弟節點，因此若樹的分 枝因數為b且最大的深度是m，則DFS所需的儲存空間只有bm 個節點。遠比BFS好很多。

DFS 的搜尋時間複雜度為O(b^m)，當目標狀態(解)有很多時 ，DFS比BFS好，但若DFS剛好選到一條路徑其解在很深很深的地方，甚至走到不歸路，那這是它的最壞狀況，這時DFS 的時間複雜度同BFS，同為O(b^d)。

由上一點我們也可得知DFS不能保證找到的解一定是最佳解。 DFS 亦不能保證若有解存在一定找得到。 因此，當一搜尋樹中有很深或無限深的路徑時，應避免使用 DFS。

Heuristic Search

This last heuristic for the driving domain is the basic idea behind most "heuristic search" techniques in artificial intelligence. If you can somehow estimate which states are "closest" to a goal state, using some numerical measure of closeness, this estimate can be used to decide which states to consider applying operators to next.

So the general idea of a heuristic search can be illustrated with this modified version of the general search algorithm:

states = make_empty_bag ()

add_to_bag (initial_state, states)

loop:

if is_empty (bag) FAIL

else:

state = remove_from_bag (bag)

if achieves_goal (new_state) SUCCEED

for each op in operators:

if op_applies (op, state)

new_state = apply_op (op, state)

add_to_bag (new_state, states)

sort_states_by_heuristic_value ()

After the new states are added to the bag, the contents of the bag are sorted so that those states for which the heuristic estimate of the distance to the goal is lowest are near the front. The state with the lowest estimate distance will be at the front and will be the next state removed from the bag and have operators applied to it.

In fact, sorting the entire contents of the bag is not always necessary. One could also implement the remove_from_bag operator to find the state whose heuristic estimate is lowest and remove that one from the bag.