Course:COGS300/Agents and Controllers

From UBC Wiki

Remembering the inputs and ASIMO as a real-world example of an agent

I know that the topic of Agents and Controllers has been iterated so many times already in the past lectures but I think that it would be nice to explain things again so as to create a better understanding of what the agent’s roles are and how the controller plays a part in an agent’s design. Also, I would like to introduce a real life example of an agent that already has been created for quite some time now. (He’s also a performer in Disneyland)


To reiterate the ideas that were already taught in class about Agents and Controllers, I made up multiple mnemonics to be able to remember what an agent is compromised of, their inputs and their interactions with the environment.


For the agent’s inputs, we have: Apes Going to a ParK Have Orange Pants 


  A being Abilities: things an agent is capable of doing
  G being Goals/Preferences: what the agent wants to achieve, it’s values and objectives
  P & K being Prior Knowledge: what an agent already knows without 
  experiencing anything yet
  H being History: the agent’s past observations of the environment
      •  O being Observations: includes current observations made by the agent
      •  P being Past experiences: includes the past observations made by the agent


Another mnemonic that explains what the agent is made of and how each part works is: Both Sam And Cam Play Cards


To explain this mnemonic, we have to put a line in between the 6 words.

Both Sam And | Cam Play Cards


The first part of each halves is what compromises the agent. The B from ‘Both’ means Body and the C from ‘Cam’ means controller. 


   The body is then made up of 
       Sensors and Actuators. The sensors are what interpret the stimuli 
       and the actuators are the parts that carry out a specific action 



    The controller takes in 
       Percepts from the body and then it sends out 
       Commands to the body to have the agent to the specific action.


Just repeat these 2 mnemonics and it would make it easier to remember what an agent does and what it’s comprised of.


A good real life example of an intelligent agent would be ASIMO. ASIMO was created by Honda to be able to help people who lack the mobility in their daily lives. According to Honda’s Design, ASIMO encompasses all of the characteristics of an agent that we described above in our mnemonic. For example, ASIMO has many different abilities but his main ‘attraction’ is his ability to walk like a human being. Honda has developed a feature called i-WALK, which is a predicted movement control technology that permits more flexible walking. Using this technology, ASIMO can shift his center of gravity in anticipation of a turn to create a more fluid motion for turning.


Since ASIMO is a humanoid robot, he has a broad set of goals that he has to complete. An example of a goal would be to deliver trays to specific locations or to hand over a tray to the specific person. (This would make it a lot easier for offices that need coffee runs every morning)


A case where ASIMO has kept prior knowledge would be his language capabilities. He has prior knowledge of a specific language like English or Japanese (depending on where ASIMO is being used) and he can use the language to form concrete sentences autonomously in response to interactions with other people.


With history, ASIMO does retain past information that he gains from previous experiences. Using his recognition technology, he can recognize faces that have been pre-registered and then he would use this knowledge to address them by their name. Although according to Honda, ASIMO can only remember up to 10 faces and hopefully they are able to remedy this in upcoming improvements.


Pertaining to ASIMO’s construction we can say that his: 

  Body:
     •	Sensors – Camera that is contained in ASIMO’s head (ASIMO has many other sensors 
       and this is just an example of one)
     •	Actuator – ASIMO has an actuator that performs like human muscles to regulate 
       his movements



  Controller:
     •	Percepts – ASIMO takes information in from his sensors, for example his camera 
     •	Commands – After taking in this perception, ASIMO responds accordingly – for example, 
       ASIMO can follow the person’s movements according to what he sees from the camera.


I have only mentioned a few of ASIMO’s features and there are many more that pertain to Agents and Controllers. The main goal here is just to understand what an Agent is and what an Agent comprises of, what it can do, etc. and also to be able to remember all the information related to Agents. (The mnemonics) The example about ASIMO is just to demonstrate how agents are applied in real life.


REFERENCES:

More information about ASIMO can be found on Honda’s website: http://asimo.honda.com/inside-asimo/

And here is a YouTube video of ASIMO performing at Disneyland: http://www.youtube.com/watch?v=rYbYP72dwYA

Retrieved from "https://wiki.ubc.ca/index.php?title=Course:COGS300/Agents_and_Controllers&oldid=261593"