Menu [toggle]

RSS Feeds [toggle]

Print

Artificial Immune System

The development of a software architecture of an artificial immune system (AIS) architecture capable of maintaining the internal state (homeostasis) for both the individual and collective SYMBRION robot.

  • Homeostasis is defined as the ability to autonomously maintain stable state in a changing environment.
    • Achieve via anomaly detection and prediction in the sensor space for the robot.

  • Homeostasis of an individual robot -> achieved through the implementation of an AIS artificial lymph node architecture.
    • Artificial lymph node provides an artificial environment for which its agents (immune cells) monitor all (sensor) inputs in order to detect as well as predict deviation from a stable state, or states, that lead to undesireble behaviour of the robot(s).
    • For example, AIS detects that a particular robotic unit has an error on board (such as a faulty actuator).
      • AIS would signal to surrounding robotic systems that that unit should not be joined into a larger robotic organism, as to do so would be detrimental to the overall organism.
    • This takes the inspiration from biological lymph nodes, and the lymphatic system by which they are connected.

  • Homeostasis of the collective organism -> A collection of SYMBRION robots, joined together to form an artificial organism creates an artificial lymphatic system.
    • Artificial lymphatic system propagates the state of one robot to other robots in the SYMBRION collective organisms.
    • Ensures the homeostasis of the collective organism is achieved and maintained.
    • This takes the inspiration from the biological lymphatic system that connects the lymph nodes.


Fig. 1. A proposed example of an immune-inspired network system



The UY group AIS wikipages consist of
  1. Artificial Immune System
  2. Overview of Immune System Biology
    • A brief biological inside on the immune system
  3. Proposed Implementation of the AIS Algorithm
    • The algorithms used for the AIS implementation
  4. Pseudocode for the Artificial Lymph Node Architecture
    • The Pseudocode for the Artificial Lymph Node Architecture to be implemented for the single SYMBRION organism
  5. Examples Scenario Dealt by AIS.




Scope of the AIS


It is important to clarify what the AIS will and will not do.

  • The AIS will not do
    • No actuator control --- The output of AIS will not suggest actuation control.
      • Interpretation of the output from AIS, could be "Warning!!! Here is the situation.".
      • To be solved by the action selection module.
    • Interact with other behaviour based controllers (at least in the short term)
    • more to be added

  • The AIS will be able to
    • DIPs under UY ToDo: M0-M12 SOFTWARE.
    • Examples Scenario Dealt by AIS.
    • Identify various abnormality in a single robotic unit
    • Once abnormality has been detected, forward messages to the action/selection system regarding the state of an individual unit
    • Identify when the organism of robots is operating outside of 'acceptable' ranges
    • When operating as an organism, share information with other AIS organisms
    • When operating in close proximity with other units, before a docking sequence begins, identify individual units that should not join the organism




Introduction to AIS


In the book Artificial Immune Systems: A New Computational Intelligence Approach [1], the following definitions of artificial immune systems (AIS) were provided:

(Starlab) defines the AIS as “data manipulation, classification, representation and reasoning methodologies which follow a biologically plausible paradigm, that of the human immune system”

(Timmis, 2000) states that “An artificial immune system is a computational system based upon metaphors of the natural immune system”

(Dasgupta, 1999) describes “Artificial immune systems are intelligent methodologies inspired by the immune system toward real-world problem solving”

The author of the same book also proposed that “Artificial immune systems are adaptive systems, inspired by theoretical immunology and observed immune functions, principles and models, which are applied to problem solving”

The definition of AIS listed above, taken from [1], were orginated from
  1. Dasgupta, D. (1999), “Information Processing Mechanisms of the Immune System,” in New Ideas in Optimisation, D. Corne, M. Dorigo and F. Glover (eds.), McGraw Hill, London, pp. 161-165.
  2. Starlab, users.pandora.be/Richard.wheeler1/ais/inn.html.
  3. Timmis, J. & Neal, M. (2000), “Investigating the Evolution and Stability of a Resource Limited Artificial Immune System,” Proc. of the Genetic and Evolutionary Computation Conference, Workshop on Artificial Immune Systems and Their Application, pp. 40-41.

A layered heuristic framework for engineering AIS was also described in [2]. The extensive description provided by [3] lists the three components of this framework. They are:
  1. Representation: How inputs are described and represented to the AIS and in what type of representation (binary, integer and float or fixed point notation)
  2. Affinity measure: Affinity measure is based on the biological measurement of strength (bind) between the receptor (immune cells receptor) and its ligand and the ideas of shape-space. Example of affinity measure are Euclidean distance, Manhattan distance, Hamming Distance, r-contiguous and r-chunk matching.
  3. Immune Algorithm: Examples of immune algorithm based biological counterpart are immune network, clonal selection, negative selection, positive selection and dendritic cell algorithm.

The three components of the heuristic framework are important in devising the AIS algorithms for its engineering application. These components, like any other AI methods, describe the functionality of the AIS in accordance to its objective.

More information on AIS is available at AISWeb The Online Home of Artificial Immune Systems at http://www.artificial-immune-systems.org




Brief Review on the Proposed Implementation


Fig. 2. indicates where AIS will sit within the higher-level controller module.


Fig. 2. Where AIS sits within the software architecture of the SYMBRION robot.


As stated above, AIS provides anomaly detection and prediction which leads to the creation of proposed output responses by the AIS (for example error codes) to the action selection module.

Again, AIS consists of:

1. Artificial lymph node maintains homeostasis of the individual SYMBRION robot.
  • Does so by first providing anomaly detection to the deviation of steady state in the input sensors values of the robot.
  • This activity is undertaken by artificial immune cells:
    • Dendritic cells -> provides anomaly detection and predction the SYMBRION robot
    • T-cells -> produces the proposed output response according the output of the Dendritics cells
    • B-cells -> provides error prediction to the robot.
  • One group of the artificial immune cells is responsible for a type of actuation variable.
    • Examples of actuation variables are: temperature control (heat up or cool down), vision (is what approaching, friend or foe), fuel (go home or recharge) and motor (speed up, slow down or maintain).
    • The output of the artificial lymph node consists of output responses from the artificial immune cells
      • These cells produced proposed output responses which help maintain the homeostasis to a single SYMBRION robot.
2. Artificial lymphatic system maintains homeostasis of a collective of SYMBRION robots
  • The proposed homeostasis control system consists of
    • homeostasis unit or one SYMBRION robot
    • homeostasis variables which evaluate the robotic control systems according to the input sensors and internal variables
    • set points that define the homeostasis region of the system.
  • An artificial lymph node is considered as one homeostasis unit of the homeostasis control system.
  • Homeostasis of the collective SYMBRION robots can be supported by the propagation of the homeostasis response of an artificial lymph node to another (artificial lymph node)
    • Thus creating an artificial lymphatic system which provides the artificial homeostasis.


How to create the AIS?

One robot is considered as one homeostasis unit. What do we need:

For homeostasis of one SYMRBION robot
  1. What are the sensor signals available to us?
  2. How the signals are represented, either raw sensor values or processed values?
  3. When do we poll for the signals (interupt or polling) and when do we provide the response?
  4. What are our response rate?

For homeostasis of collective SYMRBION robot
  1. To determine how to provide the communication between each robot in order to create the immune-inspired network architecture.
  2. What will be the signals be passed to each other (robots) to create the symbiotic organism and how will the signals be passed?
  3. How will the signals be passed to each robot?
  4. The proposed output signals: error code. Refer to Examples Scenario Dealt by AIS.

The architecture of artificial homeostasis:
  1. A homeostasis unit detects a change in the environment
  2. The homeostasis controls compares against the homeostasis variables and set points to provide the homeostasis response of the single unit.
  3. The homeostasis response of one unit is passed to its nearest neighbour of other homeostasis units.
  4. Step (1) to (3) are repeated at the neighbouring unit until a stable state is achieved and the desired response is produced for the entire immune-inspired network of the symbiotic organism.

The passing of signal between each unit can be done via the wireless communication links (ZigBee).

For the first stage, we are implementing an innate artificial immune system. This will be capable of identifying an anomaly in a single robotic unit and passing a message to the action/selection system.

For more details on the implementation of the AIS, refer to Proposed Implementation of the AIS Algorithm.

The biology behind the proposed implementation idea is in Overview of Immune System Biology


Reference
[1] Artificial Immune Systems: A New Computational Intelligence Approach, de Castro L. N. and Timmis J. (2002), Springer, 2002.
[2] Evolvable Hardware, a Fundamental Technology for Homeostasis, A. Tyrrell, J.Timmis, A.Greensted and N.Owens, 2007 IEEE Workshop on Evolvable and Adaptive Hardware proceedings, Hawaii, April 2007
[3] On Immune Inspired Homeostasis for Electronic Systems, N.Owens, J.Timmis, A.Greensted and A. Tyrrell, 6th International Conference on Artificial Immune Systems, Brazil, August 2007




Open discussion for AIS




Cut/Paste your comments/expectations about AIS here from Symbrion partners









Created by: RanBi. Last Modification: Wednesday 23 of July, 2008 14:21:08 CEST by maizura.

Menu [toggle]

Current events [toggle]

Su Mo Tu We Th Fr Sa
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31
Today

Search Wiki PageName [toggle]

RSS feed Wiki
Powered by TikiWiki CMS/Groupware