Started Labbook 2016
December 1, 2015
- Should watch this one hour lecture from the famous Neuroscientist Moser.
November 29, 2015
November 26, 2015
- Found a good lecture about Reactive architectures from Monica Nicolescu .
- Found some video's of Herbert at Jonathan Connell's webpage. Jonathan was the PhD-student who build the first subsumption robots (but not the credits).
November 24, 2015
- Read article from Roland S. Johansson on Eye-Hand Coordination in Object Manipulation.
November 23, 2015
- Found a page with an overview of Early Maze Solving Machines. The first 'Learning' machine is from 1933. See complete Scientific American 1933 article titled "Machines That Think" pdf here.
- Found the reference to actual build Braitenberg Creatures, MIT EPISTEMOLOGY AND LEARNINGMemo No. 13, 1991. Unfortunatelly, no pictures of the Lego Bricks.
November 21, 2015
- Chapter 2 of Development Robotics contains a crash-course introduction to robotics, based on Mataric's Robotic Primer.
November 20, 2015
- Checked the content of Development Robotics. The foreword seems relevant.
- The book is associated with a free MOOC from the authors (yet with registration in Italian).
- There is also another MOOC from Lyon, more based on simulated self-developing agents, with an online syllabus.
- Chapter 1 is about Growing Babies and Robots (a lot of psychology), but points to Oudeyer Playground Experiment from 2005 with an Aibo. Should search for a video.
November 12, 2015
October 28, 2015
- The book about brain-based robotics is now electronically available.
- Read chapter 1. The history is equivalent how we tell the story in the lectures. The aim is to cover current research, which could be valuable for the Fringe chapter of Arkin.
October 27, 2015
- There are 12 lectures and only 10 chapters. Luckely Arkin has some extra lectures (current choice: Developmental Learning and Swarm Robotics).
- Chapter 1 from KTH is missing, but there is big overlap between the presentations of Chapter 2.
- The presentation of Chapter 2 is 51 slides (at KTH), so could cover part I (Animal behavior) of the VSI. Part II would be Robot Behavior. Another approach would be select 12 slides from each chapter / presentation.
October 22, 2015
- Found a book from 2011 about brain-based robotics. Unfortunatelly, no inspection copies. Checked the contents, it is not a textbook, but a collection of chapters from different authors. Yet, the resource-page contains a lot of figures and videos.
I can read it in the library of the TU Delft (tag SELcb0002/DQA211 ).
September 30, 2015
September 3, 2015
- Interesting talk by Heiko Neumann on Symposium Plausible Neural Reinforcement Learning and Control. Indicates that in vision it is mostly modelled as a feedforward network, with many filtering layers improving the final association. Yet, in the brain there are many loops, partly in the layer (lateral) and from the top-layers backwards. The later gives context to the filters. To get artificial neural nets more plausible, you need attention signals (indicating which input is relevant), which is steared by the winning actions and global learning signals (when the layers should learn), corresponding with dopamine discharge. Nice vision and picture in the research report Bio-inspired Computer Vision.
August 20, 2015
August 4, 2015
- Read chapter 8 of Artificial Cognitive Systems. This chapter focus on ways to acquire knowledge (both bottom-up and top-down). Nice are the schemes of event-coding and object-action complexes.
- Read chapter 9 of Artificial Cognitive Systems. This is a short chapter about social cognition, focusing on joint actions, shared intentions, shared goals and joint attention. The comparison between Piaget's theory (a child exploring the world by first-hand experiental interaction with objects and people around it) versus Vygotsky's theory (learning by imitation and cooperation) is maybe a nice way to introduce the course.
- Overall, this was an instructive book for me, introducing cognitive theory for concepts we use in robotics. Yet, it is not a textbook; it never becomes concrete nor gives clear examples. Which could be used as concrete example is the HAMMER architecture (described in section 7.5.3). Should read the original article Hierarchical attentive multiple models for execution and recognition of actions
August 3, 2015
- Read the seventh chapter of Artificial Cognitive Systems. It describes active memory and the possibility to internal simulate the sensory consequences of motor actions with forward and inverse models.
July 9, 2015
- Read the fourth chapter of Artificial Cognitive Systems. It describes several types of autonomy, without giving useful examples from cognitive science, only some general examples as the stripes of a zebra (self-organizing pattern).
- Read the fifth chapter of Artificial Cognitive Systems. It describes three hypothesis on embodiment: conceptualization, constitution and replacement. It also gives evidence on the embodiment with some nice examples about visual attention and mirror neurons (coupled to intention, not the movement or the perception). Yet, this effidence is not used to make a clear stance for one of the hypothesis.
June 25, 2015
June 4, 2015
- Read the third chapter of Artificial Cognitive Systems. The subject is Cognitive architectures. As example, it describes a architecture from the cogitivist paradagm (SOAR), emergent paradigm (Darwin) and a hybrid architecture (ISAC). This puts the concepts introduced much more in perspective. What is missing, and where I am curious about, what the actual robots are where this architecutes are designed for.
June 1, 2015
- Read the second chapter of Artificial Cognitive Systems. It compares the cognitivist paradigm (symbolic) versus the emergent paradigm (sub-symbolic). Inside the emergent paradigm he distinguishes the connectionist approaches (neural nets), dynamic systems (phase shifts between control loops) and enactive systems, who try to learn from experience and to have sense-making in the sense that they recognize reoccuring patterns (and learn to optimize their behavior based on these patterns).
May 27, 2015
May 20, 2015
- Read chapter 10 from Autonomous Robots from G.A. Bekey. This chapter describes robot arms. Forward and inverse kinematics are mentioned, including DH-representation, but not worked out.
May 20, 2015
- Read chapter 8 and 9 from Autonomous Robots from G.A. Bekey. It describes a variety of legged robots. Nice is the example of the Phony Pony (1968), which used already FSM, contolled by wire to a micro-computer located in a neigbour building. That was 20 years before Rodney Brooks. It also tells that the usage of FSMs is quite natural to represent reoccuring patterns in gaits.
May 19, 2015
- Read chapter 7 from Autonomous Robots from G.A. Bekey. It describes a wide variety of robots (on land, in the water, in the air). Such a list can be become outdated quite fast. Did not strike me as the most important breakthroughs in mechatronics, it seems just the state-of-the-art of that age.
- Read The Ethics of Artificial Intelligence (21x cited). Boils down that it is wise to build in an ethical reasoning system, before we build superintelligent machines, because a superintelligent machine can evolve beyond our imagination and control, but with good ethics as basis we have a chance that it the ethical system evolves into human-superior niceness. The article concentrates on General AI, but points for further reading to Moral Machines: Teaching Robots Right from Wrong (276x cited).
- Oxford Press recommends the later philosophical book Guilty Robots, Happy Dogs (26x cited), from David McFarland (the same author "Animal Behavior" used by Christensen). Look inside seems to be quite complete. His earlier book from MIT is less complete.
May 17, 2015
- Read chapter 6 from Autonomous Robots from G.A. Bekey. This chapter covers learning approaches quite extensively (relatively long chapter). It describes reinforcement learnig, neural network and genetic algorithms, all illustrated by robot projects from the author himself.
- Read chapter 15 about the future developments. Most of his predictions are extensions of the state-of-the-art of that time. His longer range predictions are based on Kurzweil, Bill Joy and Hans Moravec.
May 16, 2015
- Read chapter 5 from Autonomous Robots from G.A. Bekey. This chapter should cover behavior-based robotics, but just redirects to Arkin's book for more detail. The chapter ends with a comparison of three hybrid architectures, yet those architectures (one of them OPEN-R for the AIBO) were so different, that a comparison was useless.
May 14, 2015
- Read chapter 3 and 4 from Autonomous Robots from G.A. Bekey. The 3rd chapter gives a standard overview of actuators and sensors, although somethings the link with biology is made. Chapter 4 contains quite advanced control theorie, including estimating the stability of the control loop by evaluating the phase diagram of the transfer functions. The biological inspired section pointed me to Geons (virtual sensors for 20 classes of objects, with invariance to lighting and surface characteristics).
May 13, 2015
May 4, 2015
- Read a critical review of Shanahan's Embodiment and the Inner Life. This book is the inspiration of the Ex Machina movie. Main critic seems that Shanahan correlates patterns of long-distance synchronised activity in the brain with consciousness, without any explanation why or how this two could be correlated. Read the 8 Key Themes of the book according to Shanahan himself, and this claim is theme 6. The association with raindrops in a pool makes this claim suspicious esoteric. Strange, because Prof. Shanahan is from a respected university, although it is suspicous that most of his publications are monographs.
- While analyzing Chapter 10 of Arkin, I found a collection from the Studies in Cognitive Systems series, which cited Brooks 1991 paper. Strange enough, the chapter about robotics consists of theories (Hull, Rashevsky) I never heard of. Yet, the early chapters are about Pavlov, the later ones about Wiener, so this seems to be mostly history.
- His example of a moth orienting to a light source (Jennings, 1906) is a classical Braitenburg vehicle.
- Russell created some machines with memory, to show that they could do more than pure reaction. This was 30 years ahead of cybernetics. Should be great to find some drawings of Russell's machines.
- Unfortunatelly, the series is discontinued in 2003. An interesting title is What Robots Can and Can’t Be (1992) (41x cited), who is a fan of Harnard's pretender versus thinker casus.
April 30, 2015
- Read Chapter 6 of Robot Programming. This would be a great case-study for one of the lectures. In addition, after such lecture the questions are direct applicable for the exam.
- Read Chapter 7 of Robot Programming. Clear explanation what sort of errors can be expected from each of the different sensors (concentrating on those for low budget / small mobile robot).
April 29, 2015
- Read Chapter 5 of Robot Programming. Jones is direct and honest: "I had to resort to another more complex form of arbitration (than fixed priority) on only one occasion over the course of about 15 years. Still only design exercises.
April 28, 2015
- Received the exam copy: Robot Programming: A Behavior-Based Approach by J. Jones and D. Roth. Introduction points to North-Western University : Behaviour Based Robotics (2002) by Ian Horswill, which I found already. The book is most happy with the course notes about behavior design. As the introduction says, Jones treatment is more on practical issues than on academic rigor (link to Arkin's book).
- Chapter 1 is about the sense-think-act loop. Chapter actually has 18 conceptual questions.
- Chapter 2 is about the open and closed control loops, including saturation, dead zones and backlash. The questions are still conceptual.
- Chapter 3 introduces primitive behaviors, with finite state machines (which could save information in a variable. The complexity of the design is nicely explained with a garage door example (and the Escape behavior for a mobile robot). The questions ask for instance to extend the garage door example even further (with for instance IR detection of pets).
April 24, 2015
- After several attempts (switching Aibo's, batteries, formatting sticks, etc) finally Moos started up with UU-stick (and a fresh battery).
- I could connect with pyrobot, by clicking on robot and selecting aibo.py (and specifying the ip 192.168.1.100). I could move the robot with the joystick, although it didn't had a standup behavior). Selecting the camera as device resulted in a crash (AccessInit: hash collision: 3 for both 1 and 1), which seems to indicate that both the ptz and the AiboCamera try to share the camera. Could not find the window which displays the camera-image.
April 23, 2015
- Couldn't find Tekkotsu on my Install-directory, and av-movie didn't started up any more.
- Tekkotsu is now v5.0. Yet, v2.4.1 is still available. Downloaded the ERS-7 memory-stick.
- Moos starts up and barks. I modified the wlanconf.txt (and tekkotsu.cfg), with the fixed IP 192.168.1.100, yet I could not ping Moos. Forgot to rename wlandlft.txt to wlanconf.txt! Still, no connection.
- Checked out http://svn.cs.brynmawr.edu/pyrobot/trunk (rev 2588).
- Yet, just running pyrobot didn't work. Downloaded Pyrobot Version 4.8.5. Running python pyrobot.py from C:\Programs\Python27\Lib\site-packages\pyrobot\bin started the interface. Selected as robot RoboCup, but the result is that the interface hangs.
April 22, 2015
April 16, 2015
- Found the book on scribd. Chapter 10 is only vaguely available (should join to read).
April 14, 2015
- Read Chapter 5-10 from An Introduction to robotics with Nao. Chapter 5 is a nice introduction (one or two days work), with a finite state machine to follow with the head the hand.
- Chapter 6 could be used the second day, to learn to walk and do things in parallel, but the open challenge was not challinging enough.
- Chapter 9 could be used as alternative assignments for high school kids (instead of Chapter 4), supported by Chapter 7.
- Chapter 8 could be used supportive for walking through the maze based on landmarks.
- Chapter 10 would be a nice as assignment for this course. The maze has a representation, so it would fit in the 2nd assignment 'Perception & Representation'. For the 1st assignment it would be nice if the FSM would be augmented with subsumption and inhibition.
- Looked at the assignments belonging to the Robotics Primer workbook.
- Deliberative Control has both metric and topological path planning (i.e. equivalent with the assignment of the Nao Introduction).
- Exercise2 of Hybrid Control is behavior based, but just wandering.
- Emergent Behavior has several exercises: wall following, flocking, herding. The latter ones are group behaviors. Flocking is based on BlobFinder. This is done in simulation.
- The exercise page gives an overview of all assignments, in categories.
- In Obstacle Avoidance three behaviors are combined in a BehaviorNetwork. Arbitration is done by giving commands with a priority.
- Maybe I should check Pyro, to check how they implement Behavior Networks in Python.This page even has a exercise which includes follow a wall.
- Pyro can control an Aibo ERS7 robot (with Tekkotsu stick) and has a page how to control another robot (like the Nao) by defining how it moves, translates, rotates, etc.
- Seems that the development of Pyro stopped in 2007.
- Read the Curricular materials section of AI magazine article.
- Checked UPenn. In 2009 they used Pyro and Aibo robots, in 2011 they used Lua and Nao robots. Lua stands for Lua Universal Architecture for humanoid Robotics, and looks a bit like Python.
- Actually, Luar is one of the experimental languages of Calico project, an environment intended for Robotics Education.
- It also has a textbook. Read the chapter about Behavior Control.
- Chapter 6 is about Insect-like behaviors, which describes several Braitenberg vehicles (implemented on their Miro robot). The final exercise is Corral Exiting (Maze + go the light).
- The book from Deepak Kumar is a strange combination of learning to program Python, an introduction to CS and AI, learning to control robots and accessing the Miro robot. The introduction in Python is not that bad (nor the introduction to AI), but lets concentrate.
- There is in Pyro also a module about Multiple robots, but this does only contain an example, no assignment.
- Pyro has a short introduction into Python.
- The Introduction is very handy. Yet, some of the details (robots, simulators, devices) will be out of scope.
- A nice entrance is the exercise Brain Surgery in Pyro Brains. The ptz-device is a pan-tilt-zoom camera.
April 12, 2015
- Read Chapter 10 of 'Behavior-Based Robotics'. Section 10.4 gives a list of 10 opportunities; it would be nice to see how far we have come in the last 20 years for each of this opportunities.
- It would be good if I did an analysis on which references actually had an impact and which not. In addition, which work should be added to each chapter (evolutionary approach, remove 20% and add 20%). I could write a comment on each chapter, and publish it on arXiv.org. As a second step, I could make a syllabus with an update of each chapter (based on the notes).
- Chapter 10 is short, so a nice candidate to try this approach.
April 9, 2015
April 8, 2015
- Read Chapter 6 of 'Behavior-Based Robotics'. I liked the coupling with the psychological support from Norman and Shallice Attention and Motivation model.
- Searched for robotic applications based on N-S model, found this article, where I loved Fig.1 and 2.
April 6, 2015
- Read Chapter 4 of 'Behavior-Based Robotics'. Prof. Christensen slides gives a good overview of the design principles, but does not work out the examples given in the book.
- Read Chapter 5 of 'Behavior-Based Robotics'. Introduction is nicely philosophical, with defining knowledge as information which both applicable and should be able to make predictions. Clearly, perception and knowledge representation has made big changes in 10 years. Prof. Christensen slides only partly cover the examples given in the book.
April 3, 2015
- Read Chapter 3 of 'Behavior-Based Robotics'. Prof. Christensen slides are rather short summaries of all the details given in the book, I would make the presentation twice as long.
April 2, 2015
- Read Chapter 38 of 'Springer Handbook of Robotics'. Maja mainly referstot Arkin's book (and her own thesis), yet this chapter is great to give overview what happened after Arkin's book.
- Maja indicates in further reading Brook's Cambrian Intelligence.
- Should check if Maja or Francois have written another textbook.
- Also no textbook of Prof. Christensen; the Behavior-Based Robotics course is taught for the last time in 2004 (never at Georgia Tech).
March 31, 2015
- Read Chapter 1 and 2 of 'Behavior-Based Robotics'. Chapter 2 explains the cognitive, psychological and ethological basis of behaviors.