Labbook SI4MS

12 December

• Read the Dutch ITS plan 2013-2017.
• Found a tutorial about Dutch GeoDATA, which can be loaded in the GIS Arcmap.
• Another nice guide is the about the web-based delivery of open source European geodata, Inspire FOSS.

4 October

• Searched for the LCGL library, as mentioned in Autonomous Vehicles in City Traffic and A perception-driven autonomous urban vehicle. Checked out Camunits library. The code makes use of OpenGL, but I couldn't find any reference to lcm. Inside lcm, I couldn't find reference to OpenGL. Inside the lcmtypes, I only found procman_process_t, which can execute an command-line command (yet, this seems a bit heavy for a rendering).

2 October

• Finished reading the article of the MIT Darpa team about their architecture.
• Read the Communication Architecture chapter of WP1.
• Read the sheets of Erlangen. WAVE is new DSRC term, RSU is a Wave Acces Point and OBU is a Wave Station. The LCM broadcast could at best be a Wave aware UDP application (with help of an ISO 21210 convergence layer).
• Found another QFree presentation, with FOAM as end2end framework above CALM, and POMALOR and COMO as two applications (map services and vehicle monitoring resp) on top of FOAM (as part of CVIS). FOAM seems to be implemented as Java based components exchanging data in fixed exchange formats).
• The Magic team used OM1P Mesh Routers for short-range communication, the predessor of the OM2P Mesh Routers.

26 September

• Writing on the TechReport.
• Found the lcm-dataset page from the MIT Urban Challenge team.
• Start reading the article of the MIT Darpa team about their architecture.

19 September

• Checked the code out at pc-udk. Opened the solution with Visual Studio 11. Projects were converted without problems; program was build without any warnings.
• Code worked fine (didn't check with LcmSpy, but output seems OK). RoboCup2012-Tryout gave warnings about Envy. Copied the UDKGame files from Mexico. Problem solved. make_clean gave warning about radar. Didn't looked into that. Running keeps on complaining that the scripts are outdated (also after a make). Yet, no make screen pops up. Tried to change the command from 32bits to 64bits, but that didn't help.

18 September

• Made the documentation to be included with the LCM USARSim Proxy distribution.
• Cleaned up the code and published the code to sourceforge.
• Made a summary of the documentation available at the USARSim Manual.

17 September

• Read the Architectures chapter from the Robotics Handbook. JAUS is mentioned as architecture component (mixture of client-server and publish-subscribe), while RCS is mentioned as a multitiered architecture. No mentioning of LCM, URBI or ROS.
• Made package ready for publication. VS2005 cannot handle .Net Framework 4, so had to manually reduce the Tools reference to v3.5. Couldn't find any Framework64 Systems.dll, so switched back to Any PC configuration. Tested code, seems to work. Package is now ready for publication, only have to clean up code and add documentation.

16 September

• Looked into the Multicast UDP issue. According to twisted internet documentation, the multicast address is a group address (in their case "228.0.0.5").
• According to a multicast tutorial: 'This reserved range consists of addresses from 224.0.0.0 to 239.255.255.255. However, the multicast addresses from 224.0.0.0 to 224.0.0.255 are reserved for multicast routing information;'. Setting TTL to 128 means that the messages are broadcasted to whole continent! The turorial also contains a 'Hello world' example.
• Modified my C#-code so that it publishes on 'udpm://239.255.76.67:7667?ttl=1' instead of a singleton publishing on the localhost. Both nb-udk and abeel were in the same DNT subnet (192.168.1.82 and 83) and I could see the published data on the Linux-machine with command java -ea -server lcm.spy.Spy -l udpm://239.255.76.67:7667?ttl=1:
  
• Made some code to broadcast the image data. It is a bit difficult to get the length and width from a jpeg-stream (luckely they are fixed). Nothing visible when I call the spy (also no Laser of Pose data).

2 September, 2012

• Read the Tech Report about LCM. Lcm communication architecture is compared with JAUS, Player, Carmen and ROS.
• As described in Richard's Project Report, lcm can also easily coupled to USARSim.

6 March 2012

• Read the article 'Studying complex adaptive systems' by J.H. Holland. There are three open challenges:
  1. The prediction of lever points in complex adaptive systems.
  2. The hierarchical organizations of boundaries inside a system.
  3. Open-ended evolution of experience.
• Lets concentrate on the lever points. This fits into Bernard van Veelen's Emergence Control. The CAS-wiki relates them with Frozen accidents or Choke points, although those seemed to focus on events that influence the system in a negative way. Holland's definition is more neutral, more equivalent with a mutation in an evalutionary system. Anyway, the CAS-wiki directly relates frozen accidents with emergence.
• Read Donella Meadows' article about leverage points. This was not about events, but on places to intervene in a system. She made a list of those place:
  • Constants, parameters
  • buffer sizes
  • connections between buffers
  • delays in connections
  • strength of corrective feedback loops (PI)
  • gains of reenforcing feedback loops (PID)
  • structure of information flows (creating non-physical shortcuts)
  • rules of the system
  • power to change, evolve, adapt
  • goals of the system
  • paradigms (fairness, culture)
  • power to change culture
  The list was experience and example based, no theory or evidence was given. The list was ordered in effectiveness, but as caveat she said that increased effectiveness also means that there are more strong conservative powers to resist change on those places.
• The wiki-page about Emergence quite elaborated. Nice idea is to connect the type of emergence through the lense of implementation (the levels of abstraction). Note that Holland suggested that 2 levels were already enough for intelligent behavior).
• A self-organization paper gives a good concise overview of the field. According to Francis Heylighen: The concept of self-organization was first proposed by the cyberneticist W. Ross Ashby (1) in the 1940s and developed among others by his colleague Heinz von Foerster (2). ... During the 1960s and 1970s, the idea was picked up by physicists and chemists studying phase transitions and other phenomena of spontaneous ordering of molecules and particles. In the 1980s, this tradition cross-fertilized with the emerging mathematics of non-linear dynamics and chaos, producing an investigation of complex systems that is mostly quantitative, mathematical, and practiced by physicists. However, the same period saw the appearance of a parallel tradition of 'complex adaptive systems' (5), associated with the newly founded Santa Fe Institute for the sciences of complexity, that is closer in spirit to the cybernetic roots of the field. Building on the work of John Holland, Stuart Kauffman, Robert Axelrod, Brian Arthur and other SFI associates, this approach is more qualitative and rooted in computer simulation. It took its inspiration more from biology and the social sciences than from physics and chemistry, thus helping to create the new disciplines of artificial life and social simulation. It seems that this history is based on (5) Waldrop, M. M. Complexity: The Emerging Science at the Edge of Order and Chaos, London: Viking, 1992.
  Also note the list of further reading at the end of the article.

5 January 2012

• Read Consciosusness in Cognitive Architectures. Chapter 3 refers to Klir's General System Theory. Chapter 5 contains the PERMIS evaluation criteria for intelligent systems, which are very applicable.

Previous Labbook

• Labbook 2011.