Robotics control is a hard task because of the combinatorial explosion which references to a huge... more Robotics control is a hard task because of the combinatorial explosion which references to a huge game state. There are endless amount of possible sensor readings and it is impossible to plan actions into the future. With the help of an abstraction mechanism the state space gets converted into a low complexity representation which is mostly a communication board providing words from a controlled vocabulary. These words are forming a new state space which is much smaller than the original sensor space from the robot. The following paper describes a scene parser which translates back and forth between sensor readings and a word table which allows to control robots. It's not about the control problem itself but the interaction between human and robot stands in the center of attention. A working prototype is available under https://youtu.be/S9xCXpNDYRg
Bis ungefähr dem Jahr 2000 wurden Roboter mit zwei zentralen Prinzipien programmiert: der Roboter... more Bis ungefähr dem Jahr 2000 wurden Roboter mit zwei zentralen Prinzipien programmiert: der Roboter agierte autonom und er war nicht in der Lage natürliche Sprache zu verstehen. Diese Prinzipien wurden als Erfolgsgarant für ernsthafte Künstliche Intelligenz verstanden. Nur damit ließen sich mathematisch korrekte Algorithmen entwickeln, die die Anforderungen in der Prozessautomatisierung erfüllten. Das vorliegende wissenschaftliche Paper beschreibt ein gegensätzliches Modell. Im Zentrum steht das Symbol grounding problem bei dem Sensor Signale in natürliche Sprache übersetzt werden. Es kommen keine bestimmten Algorithmen zum Einsatz sondern es wird eine Notation beschrieben in Form eines kontrollierten Vokabulars. Diese Minisprache verleiht den Sensordaten Bedeutung. Die Übersetzung von Mausgesten in ein natürlich-sprachliches Gamelog ist die Aufgabe des Parsers. Ein Video zur Simulation eines Küchenroboters ist vorhanden auf https://youtu.be/Qj_fBJWM97w
In the past, the robot's brain was located in a certain algorithm which is executed on the CPU. T... more In the past, the robot's brain was located in a certain algorithm which is executed on the CPU. The estimated runtime for dedicated AI np-hard algorithm was high. It was impossible to determine the optimal action and most robotics projects have failed. The following paper introduces sensor driven robotics which is working with data structures stored in the memory. The raw sensor data are converted into detected events similar what a soccer referee is doing. Events are stored in a gantt chart which is a temporal model and high level event sequences are recognized. The overall pipeline is addressing the symbol grounding problem.
Creating advanced robot control systems can be realized with a reward function. The principle is ... more Creating advanced robot control systems can be realized with a reward function. The principle is known from the potential field path planning method and gets adapted to robotics in general. The idea is to create a continuous cost function by hand and assign each state of the game a numerical cost value. This pre-processing step converts complicated games into easier ones. If the potential field for the action space is known, it is pretty easy to find the optimal action.
The wikipedia draft space is hard to predict for newbie authors. This makes it hard to submit new... more The wikipedia draft space is hard to predict for newbie authors. This makes it hard to submit new draft articles into the main space. The rejection rate is high. More than 50% of all the draft submissions are not allowed to enter the main space. This paper explains the reason why and gives advice how to increase the success probability. Tags: Wikipedia Contents
The main bottleneck in the Wikipedia project isn't the Visual editor or the creation of a user ac... more The main bottleneck in the Wikipedia project isn't the Visual editor or the creation of a user accounts, but it's the editing workflow itself. To speed up the writing, the user has to join a Wikiproject, create keypoints but not full sentences, and doing so for many articles in parallel. This will balance the needs of the user to use Wikipedia as a blogging tool with the objectives of admins to ensure a high quality standard.
The average timelag between introducing a new security problem in the upstream and until it's fix... more The average timelag between introducing a new security problem in the upstream and until it's fixed by downstream security experts is around 6.9 years. During this time span the software has vulnera-bilities. It's the obligation of the Debian distribution to educate the normal user about computer security in general. Also, Debian has to provide information which security problem was solved and how long did it take.
Analyzing the inner working of the largest encyclopedia in the world, is a complicated task. The ... more Analyzing the inner working of the largest encyclopedia in the world, is a complicated task. The mediawiki software itself is an easy to understand part, but describing how the users are interacting with the software to create content is the more demanding task. The best practice method in interacting with wikipedia is to do a fulltext in the backlog. This allows to identify a specific task which is interesting for the user. He creates an edit to solve the task and submits it to the encyclopedia. If the edit wasn't made undo, the user is proud and feels accepted.
Task and motion planning separates a robotics domain into two layers. The high-level layer is des... more Task and motion planning separates a robotics domain into two layers. The high-level layer is described in a pddl like formalization while the lower layer is equal to low level behavior which is formalized in motion commands. A well known datastructure for storing the knowledge is called petri-net which is a graph based semantic network which holds linked lists and can update them easily. In the following paper some techniques are explained used in the domain of robotics and computer animation to solve the task and motion planning problem.
Vor einiger Zeit habe ich ein 35 Seiten Dokument geschrieben, was die Geschichte der Robotik erlä... more Vor einiger Zeit habe ich ein 35 Seiten Dokument geschrieben, was die Geschichte der Robotik erläutert. Darin gibt es ein eigenes Kapitel über die MIT/Utah Hand sowie ein wenig überraschend auch eines über die Geschichte der Siggraph-Konferenz, welche als Ideengeber für viele Projekte der Künstlichen Intelligenz diente. Der Text ist auf Deutsch geschrieben. Abbildungen sind übrigens keine enthalten; es ist eine reine Bleiwüste. Dafür aber in ordentlichem LaTeX formatiert und mit einem hübschen Literaturverzeichnis.
Forth hat keine Standard-Bibliotheken und die Hardware ist nicht compiler-friendly. Das macht For... more Forth hat keine Standard-Bibliotheken und die Hardware ist nicht compiler-friendly. Das macht Forth zu einer Toy-Language. Man kann damit keine größeren Projekte realisieren und für die Robotik ist Forth ungeeignet. Will man in Forth programmieren, fallen extrem hohe Kosten an.
Der Verlauf eines Softwareprojektes wird tagebuchartig beschrieben. Es wird als Programmiersprach... more Der Verlauf eines Softwareprojektes wird tagebuchartig beschrieben. Es wird als Programmiersprache Python verwendet, wobei objektorientierte Entwicklung einen hohen Stellenwert besitzt. Ergänzend kommt das Versionsverwaltungssystem git zum Einsatz. Der Text ist kein Lehrbuch, sondern ein Projekt-Log-Buch. Demzufolge gibt es viel Sourcecode, Berichte über gescheiterte Versuche und Sackgassen aus denen man hoffentlich etwas lernen kann.
Ausgehend von der Idee eines Head-up Displays lässt sich die Domäne des autonomen Fahrens modelli... more Ausgehend von der Idee eines Head-up Displays lässt sich die Domäne des autonomen Fahrens modellieren. Anders als bei Behavior Trees wird das Knowledge-Modell nicht direkt zur Ausführung gebracht sondern der menschliche Fahrer übernimmt die Steuerung des Fahrzeuges.
Die üblicherweise diskutierte Frontlinie “Microsoft vs. Linux” ist ungenau. Eigentlich verläuft d... more Die üblicherweise diskutierte Frontlinie “Microsoft vs. Linux” ist ungenau. Eigentlich verläuft der Graben durch die Linux Community selbst. Einmal gibt es die Firma Red Hat welche ein Monopol auf den Upstream besitzt und als Gegner dazu gibt es Red Hat Clone wie Debian, welche mit Absicht inkompatibel sind. Red Hat verteidigt sich an mehreren Fronten zugleich: einmal muss er versuchen, Kunden zu überzeugen, dann wird der Konzern von anderen LInux Distributionen attackiert und als Krönung kommt Gegenwind von den Linux-Medien wie Heise Online. Red Hat ist damit das Opfer einer großangelegten Verschwörung. Diese zielt darauf ab, die Deutungshoheit über freie Software zu übernehmen.
Die komplette Linux Community hat sich gegen Red Hat verschworen. Ubuntu, ArchLinux und Debian fü... more Die komplette Linux Community hat sich gegen Red Hat verschworen. Ubuntu, ArchLinux und Debian führen einen Krieg gegen RHEL. Sie wollen Red Hat kaputt machen. Beteiligt an dieser Anti-Redhat Verschwörung sind auch die Medien, also die Linux Zeitschriften und dezidierte Linux-Podcasts. Dort wird überall das Märchen verbreitet, dass GNU Debian Linux die Zukunft wäre, und das Red Hat nur Geld verdienen will.
Tagebuchartig wird die Erstellung eines Robot-Control-Systems in der Programmiersprache Python er... more Tagebuchartig wird die Erstellung eines Robot-Control-Systems in der Programmiersprache Python erläutert. Das Controller-Design orientiert sich an dem Interactive Animation Control Paradigma. Zwischendurch gab es allerhand Schwierigkeiten zu überwinden was nicht immer gelang.
Klassisches akademisches Publishing ist so langsam wie die königliche Postkutsche. Es gilt allerh... more Klassisches akademisches Publishing ist so langsam wie die königliche Postkutsche. Es gilt allerhand Traditionen zu beachten und das Schreiben eines Dokumentes erfolgt mit einem Federkelch, der be-hutsam in Tinte getaucht wird. Die Pferde kommen angesichts der schlecht ausgebauten Wege nur langsam voran und müssen aufwen-dig gestriegelt werden. Bis die Post befördert ist, vergehen Wochen, ja Monate, aber das stört niemanden weil die Menschen auf dem Lande viel Zeit haben.
Echtzeitprogrammierung stellt nur geringe Anforderungen an die CPU. Im Normalfall muss nur alle 1... more Echtzeitprogrammierung stellt nur geringe Anforderungen an die CPU. Im Normalfall muss nur alle 1 Sekunde ein Steuerimpuls generiert werden. Das ist langsamer als die Bildwiederholfrequenz in einer Game-Engine. Schwierig ist jedoch die Modellierung eines Echtzeitsystems. Formal handelt es sich um ein Hybrid-System bestehend aus einer Finite-State-Machine die um analoge Differentialgleichungen ergänzt wurde.
RRT ist ein Brute-Force-Algorithmus, der einen Baum als Datenstruk-tur nutzt um vorhandene Pfade ... more RRT ist ein Brute-Force-Algorithmus, der einen Baum als Datenstruk-tur nutzt um vorhandene Pfade zu speichern. Das Verfahren ist uni-versell einsetzbar als Pfadplanner, Motion Planner und zur Bestim-mung von inverser Kinematik. Um eine gleichmäßige Ausbreitung des Graphen zu erreichen wird ein Zufälliger Punkt außerhalb des Gra-phen verwendet und dafür der nächst gelegene Node bestimmt. Hier gibt es Überschneidungen zum PRM Verfahren.
Robotics control is a hard task because of the combinatorial explosion which references to a huge... more Robotics control is a hard task because of the combinatorial explosion which references to a huge game state. There are endless amount of possible sensor readings and it is impossible to plan actions into the future. With the help of an abstraction mechanism the state space gets converted into a low complexity representation which is mostly a communication board providing words from a controlled vocabulary. These words are forming a new state space which is much smaller than the original sensor space from the robot. The following paper describes a scene parser which translates back and forth between sensor readings and a word table which allows to control robots. It's not about the control problem itself but the interaction between human and robot stands in the center of attention. A working prototype is available under https://youtu.be/S9xCXpNDYRg
Bis ungefähr dem Jahr 2000 wurden Roboter mit zwei zentralen Prinzipien programmiert: der Roboter... more Bis ungefähr dem Jahr 2000 wurden Roboter mit zwei zentralen Prinzipien programmiert: der Roboter agierte autonom und er war nicht in der Lage natürliche Sprache zu verstehen. Diese Prinzipien wurden als Erfolgsgarant für ernsthafte Künstliche Intelligenz verstanden. Nur damit ließen sich mathematisch korrekte Algorithmen entwickeln, die die Anforderungen in der Prozessautomatisierung erfüllten. Das vorliegende wissenschaftliche Paper beschreibt ein gegensätzliches Modell. Im Zentrum steht das Symbol grounding problem bei dem Sensor Signale in natürliche Sprache übersetzt werden. Es kommen keine bestimmten Algorithmen zum Einsatz sondern es wird eine Notation beschrieben in Form eines kontrollierten Vokabulars. Diese Minisprache verleiht den Sensordaten Bedeutung. Die Übersetzung von Mausgesten in ein natürlich-sprachliches Gamelog ist die Aufgabe des Parsers. Ein Video zur Simulation eines Küchenroboters ist vorhanden auf https://youtu.be/Qj_fBJWM97w
In the past, the robot's brain was located in a certain algorithm which is executed on the CPU. T... more In the past, the robot's brain was located in a certain algorithm which is executed on the CPU. The estimated runtime for dedicated AI np-hard algorithm was high. It was impossible to determine the optimal action and most robotics projects have failed. The following paper introduces sensor driven robotics which is working with data structures stored in the memory. The raw sensor data are converted into detected events similar what a soccer referee is doing. Events are stored in a gantt chart which is a temporal model and high level event sequences are recognized. The overall pipeline is addressing the symbol grounding problem.
Creating advanced robot control systems can be realized with a reward function. The principle is ... more Creating advanced robot control systems can be realized with a reward function. The principle is known from the potential field path planning method and gets adapted to robotics in general. The idea is to create a continuous cost function by hand and assign each state of the game a numerical cost value. This pre-processing step converts complicated games into easier ones. If the potential field for the action space is known, it is pretty easy to find the optimal action.
The wikipedia draft space is hard to predict for newbie authors. This makes it hard to submit new... more The wikipedia draft space is hard to predict for newbie authors. This makes it hard to submit new draft articles into the main space. The rejection rate is high. More than 50% of all the draft submissions are not allowed to enter the main space. This paper explains the reason why and gives advice how to increase the success probability. Tags: Wikipedia Contents
The main bottleneck in the Wikipedia project isn't the Visual editor or the creation of a user ac... more The main bottleneck in the Wikipedia project isn't the Visual editor or the creation of a user accounts, but it's the editing workflow itself. To speed up the writing, the user has to join a Wikiproject, create keypoints but not full sentences, and doing so for many articles in parallel. This will balance the needs of the user to use Wikipedia as a blogging tool with the objectives of admins to ensure a high quality standard.
The average timelag between introducing a new security problem in the upstream and until it's fix... more The average timelag between introducing a new security problem in the upstream and until it's fixed by downstream security experts is around 6.9 years. During this time span the software has vulnera-bilities. It's the obligation of the Debian distribution to educate the normal user about computer security in general. Also, Debian has to provide information which security problem was solved and how long did it take.
Analyzing the inner working of the largest encyclopedia in the world, is a complicated task. The ... more Analyzing the inner working of the largest encyclopedia in the world, is a complicated task. The mediawiki software itself is an easy to understand part, but describing how the users are interacting with the software to create content is the more demanding task. The best practice method in interacting with wikipedia is to do a fulltext in the backlog. This allows to identify a specific task which is interesting for the user. He creates an edit to solve the task and submits it to the encyclopedia. If the edit wasn't made undo, the user is proud and feels accepted.
Task and motion planning separates a robotics domain into two layers. The high-level layer is des... more Task and motion planning separates a robotics domain into two layers. The high-level layer is described in a pddl like formalization while the lower layer is equal to low level behavior which is formalized in motion commands. A well known datastructure for storing the knowledge is called petri-net which is a graph based semantic network which holds linked lists and can update them easily. In the following paper some techniques are explained used in the domain of robotics and computer animation to solve the task and motion planning problem.
Vor einiger Zeit habe ich ein 35 Seiten Dokument geschrieben, was die Geschichte der Robotik erlä... more Vor einiger Zeit habe ich ein 35 Seiten Dokument geschrieben, was die Geschichte der Robotik erläutert. Darin gibt es ein eigenes Kapitel über die MIT/Utah Hand sowie ein wenig überraschend auch eines über die Geschichte der Siggraph-Konferenz, welche als Ideengeber für viele Projekte der Künstlichen Intelligenz diente. Der Text ist auf Deutsch geschrieben. Abbildungen sind übrigens keine enthalten; es ist eine reine Bleiwüste. Dafür aber in ordentlichem LaTeX formatiert und mit einem hübschen Literaturverzeichnis.
Forth hat keine Standard-Bibliotheken und die Hardware ist nicht compiler-friendly. Das macht For... more Forth hat keine Standard-Bibliotheken und die Hardware ist nicht compiler-friendly. Das macht Forth zu einer Toy-Language. Man kann damit keine größeren Projekte realisieren und für die Robotik ist Forth ungeeignet. Will man in Forth programmieren, fallen extrem hohe Kosten an.
Der Verlauf eines Softwareprojektes wird tagebuchartig beschrieben. Es wird als Programmiersprach... more Der Verlauf eines Softwareprojektes wird tagebuchartig beschrieben. Es wird als Programmiersprache Python verwendet, wobei objektorientierte Entwicklung einen hohen Stellenwert besitzt. Ergänzend kommt das Versionsverwaltungssystem git zum Einsatz. Der Text ist kein Lehrbuch, sondern ein Projekt-Log-Buch. Demzufolge gibt es viel Sourcecode, Berichte über gescheiterte Versuche und Sackgassen aus denen man hoffentlich etwas lernen kann.
Ausgehend von der Idee eines Head-up Displays lässt sich die Domäne des autonomen Fahrens modelli... more Ausgehend von der Idee eines Head-up Displays lässt sich die Domäne des autonomen Fahrens modellieren. Anders als bei Behavior Trees wird das Knowledge-Modell nicht direkt zur Ausführung gebracht sondern der menschliche Fahrer übernimmt die Steuerung des Fahrzeuges.
Die üblicherweise diskutierte Frontlinie “Microsoft vs. Linux” ist ungenau. Eigentlich verläuft d... more Die üblicherweise diskutierte Frontlinie “Microsoft vs. Linux” ist ungenau. Eigentlich verläuft der Graben durch die Linux Community selbst. Einmal gibt es die Firma Red Hat welche ein Monopol auf den Upstream besitzt und als Gegner dazu gibt es Red Hat Clone wie Debian, welche mit Absicht inkompatibel sind. Red Hat verteidigt sich an mehreren Fronten zugleich: einmal muss er versuchen, Kunden zu überzeugen, dann wird der Konzern von anderen LInux Distributionen attackiert und als Krönung kommt Gegenwind von den Linux-Medien wie Heise Online. Red Hat ist damit das Opfer einer großangelegten Verschwörung. Diese zielt darauf ab, die Deutungshoheit über freie Software zu übernehmen.
Die komplette Linux Community hat sich gegen Red Hat verschworen. Ubuntu, ArchLinux und Debian fü... more Die komplette Linux Community hat sich gegen Red Hat verschworen. Ubuntu, ArchLinux und Debian führen einen Krieg gegen RHEL. Sie wollen Red Hat kaputt machen. Beteiligt an dieser Anti-Redhat Verschwörung sind auch die Medien, also die Linux Zeitschriften und dezidierte Linux-Podcasts. Dort wird überall das Märchen verbreitet, dass GNU Debian Linux die Zukunft wäre, und das Red Hat nur Geld verdienen will.
Tagebuchartig wird die Erstellung eines Robot-Control-Systems in der Programmiersprache Python er... more Tagebuchartig wird die Erstellung eines Robot-Control-Systems in der Programmiersprache Python erläutert. Das Controller-Design orientiert sich an dem Interactive Animation Control Paradigma. Zwischendurch gab es allerhand Schwierigkeiten zu überwinden was nicht immer gelang.
Klassisches akademisches Publishing ist so langsam wie die königliche Postkutsche. Es gilt allerh... more Klassisches akademisches Publishing ist so langsam wie die königliche Postkutsche. Es gilt allerhand Traditionen zu beachten und das Schreiben eines Dokumentes erfolgt mit einem Federkelch, der be-hutsam in Tinte getaucht wird. Die Pferde kommen angesichts der schlecht ausgebauten Wege nur langsam voran und müssen aufwen-dig gestriegelt werden. Bis die Post befördert ist, vergehen Wochen, ja Monate, aber das stört niemanden weil die Menschen auf dem Lande viel Zeit haben.
Echtzeitprogrammierung stellt nur geringe Anforderungen an die CPU. Im Normalfall muss nur alle 1... more Echtzeitprogrammierung stellt nur geringe Anforderungen an die CPU. Im Normalfall muss nur alle 1 Sekunde ein Steuerimpuls generiert werden. Das ist langsamer als die Bildwiederholfrequenz in einer Game-Engine. Schwierig ist jedoch die Modellierung eines Echtzeitsystems. Formal handelt es sich um ein Hybrid-System bestehend aus einer Finite-State-Machine die um analoge Differentialgleichungen ergänzt wurde.
RRT ist ein Brute-Force-Algorithmus, der einen Baum als Datenstruk-tur nutzt um vorhandene Pfade ... more RRT ist ein Brute-Force-Algorithmus, der einen Baum als Datenstruk-tur nutzt um vorhandene Pfade zu speichern. Das Verfahren ist uni-versell einsetzbar als Pfadplanner, Motion Planner und zur Bestim-mung von inverser Kinematik. Um eine gleichmäßige Ausbreitung des Graphen zu erreichen wird ein Zufälliger Punkt außerhalb des Gra-phen verwendet und dafür der nächst gelegene Node bestimmt. Hier gibt es Überschneidungen zum PRM Verfahren.
Die automatisierte Steuerung von Robotern scheiterte in der Vergangenheit häufig an der enormen G... more Die automatisierte Steuerung von Robotern scheiterte in der Vergangenheit häufig an der enormen Größe des Zustandsraumes. Der Grund sind hochauflösende Sensoren in Verbindung mit Servomotoren, die numerische Werte als Eingabe verwenden. Selbst moderne Computer sind überfordert, den Spielbaum für derartige lowlevel Zustandsräume zu bestimmen, geschweige denn mehrere Schritte in die Zukunft zu planen. Der Ausweg aus dem Dilemma ist die Projektion des ursprünglichen Zustandsraumes in einen abstrakten kognitiven Zustandsraum. Dieses Mapping ist auch als Symbol grounding Problem bekannt und bedeutet die Transformation von Sensordaten in eine symbolische Schrift. Es handelt sich um ein Kodierungsproblem zur Repräsentation der Realität in einem Roboter. Das Erzeugen einer linguistischen Grammatik ist Teil der folgenden Ausarbeitung. Ferner wird am Beispiel eines Autorennspiels demonstriert, wie eine natürlichsprachliche Beschreibung der Ist Situation aussieht.
Classical computer science assumes a closed system which runs an algorithm. The algorithm has to ... more Classical computer science assumes a closed system which runs an algorithm. The algorithm has to find a trajectory for a mathematical optimization problem. Unfortunately, this popular approach generates np hard problems which can't be solved on a computer in realtime because of the large state space. Even possible future computers will become too slow to analyze a game tree which has millions of millions possible situations. Graph search algorithms including alpha beta pruning can't handle the situation very well because of the missing problem decomposition and a lack of expert knowledge. The answer to the state space explosion is a dialogue based communication with an open system. A human speaker sends grammar based messages to a computerized hearer. These language based heuristics are described in the following paper including two practical examples which are the game of snake and a quadruped walking robot.
Classical Turing machines are controlled by algorithms, not with input from the environment. The ... more Classical Turing machines are controlled by algorithms, not with input from the environment. The program itself, executed by the Turing machine, determines the next action. Unfortunately, there is no algorithm available to solve np hard motion planning problems and more advanced problems from robotics remain unsolved. The gap can be closed with so called Oracle Turing Machines who are communicating with an external authority. No algorithm is required for a communication but a vocabulary of words. The oracle gives a command e.g. "move north" and the Turing machine is executing the command similar to a game engine. Such kind of distributed intelligence can tackle np hard problems. The absence of a control algorithm in robotics is a novel technique and it requires a new sort of problem understanding. Instead of solving an existing well defined problem with a computer program the idea is to create a dataset with a speaker to hearer dialogue and convert the dataset into a communication protocol.
The core element of the symbol grounding problem is the instruction following task which is a dia... more The core element of the symbol grounding problem is the instruction following task which is a dialogue based language game. The speaker gives a sequence of commands which have to be interpreted by the hearer. The dialog allows to solve complex problems from robotics which can't be addressed with backtracking algorithms from the past. From a programming perspective, the vocabulary of an instruction following game has much in common with the opcodes in Assembly language used to program computers on the lowest level. The idea is that a high level input language gets converted into a low level machine language. This transformation is done by an interpreter which is a grounding mechanism. In addition to the previous two articles about the symbol grounding problem, this part III gives an overview about the origin and how it may help to build future robots.
The dominant challenge in robotics is the large state space. Any possible algorithm which travers... more The dominant challenge in robotics is the large state space. Any possible algorithm which traverses the entire state space will become too slow even on supercomputing hardware. Overcoming the bottleneck can be realized with guided instructions which are resulting into a speaker hearer interaction. The high level commands from the speaker are interpreted by the hearer which is the robot. Even if the concept sounds a bit uncommon, it makes sense to investigate it in detail because it allows to solve complex robotics problem with natural language. After a short introduction into the last AI Winter which went until the year 1992, the dominant part of the following paper is dedicated to the instruction following problem. A simple maze game is taken as an example to implement a working prototype in Python which allows to control a robot swarm with basic commands.
Robotics in the past was described as algorithm centric problems. The idea was that a certain com... more Robotics in the past was described as algorithm centric problems. The idea was that a certain computer program is able to determine the optimal action. This decision is made, according to the computational paradigm, by the CPU which is running some sort of intelligent program. The result of this understanding was to locate the intelligence inside the Turing machine which gets executed on a physical computer. In contrast, the symbol grounding problem assumes that the bottleneck is located in man to machine communication which is realized with a protocol. The protocol gets implemented with semantic tagging which maps keywords to certain game states. The sum of all keywords are forming a cognitive map which can be extended to an activity language and possible voice commands. The shared vocabulary between man and machine has much in common with a text adventure. The following paper introduces the symbol grounding problem for the example of a kitchen robot and the pong videogame.
Social networks are referencing similar to annotated bibliographies to existing sources outside t... more Social networks are referencing similar to annotated bibliographies to existing sources outside the system. A typical entry in a bibliography and a social network as well contains of date, URL and comment. Creating such an entry is a cheap operation and can be realized by a user in under 10 minutes. This results into a large amount of postings per user and a high acceptance rate among all users. The interesting point is, that around a bibliography a dedicated community can be realized which are posting their favorite links to the system to get upvotes from other users. Such kind of group oriented evaluation website can solve the peer review issue in the Open Science movement.
Customer orientation is equal to provide high quality academic content to everybody. According to... more Customer orientation is equal to provide high quality academic content to everybody. According to the business model "Author pays" the reader is king and dominates the relationship. The transformation into an Open Access ecosystem follows similar rules like in the Web 2.0 age, which is about user-generated content by amateurs. This will lower the entry barrier into the ivory tower and allows the users to participate frequently.
Peer review can be realized with a low quality journal. Peer review means simply an interaction b... more Peer review can be realized with a low quality journal. Peer review means simply an interaction between authors and readers. The idea is that the authors doesn't publish everything they have written and as a result the amount of information is reduced for the reader. A voting system is needed as an intermediate.
Artificial Intelligence can be realized with AI planning. That is a sub-discipline about creating... more Artificial Intelligence can be realized with AI planning. That is a sub-discipline about creating plans to bring a system into a goal state. In traditional planning only shallow models are used, for example the A* pathplanner. With a hierarchical extension, it is possible to use planning also for complex domains. The overall system is called in the literature an agent based blackboard architecture which has the aim of grounding natural language into an action model.
Programming an artificial intelligence is easy, if exhaustive search is possible to find the best... more Programming an artificial intelligence is easy, if exhaustive search is possible to find the best node in the game tree. The program tries out all possibilities and searches for a path to win the game. Most robotics domains have the problem, that the state space is very huge and a brute-force search is not possible on slow computer hardware. The answer to the problem is called heuristic search. A heuristic is a way to shorten the search, this is done with abstraction. The overall domain is subdivided into smaller simulations which are predicting different aspects of the game. The simulation has different layers to increase the abstraction further.
Open Access publishing has nothing to do with internet based technology or fiber-optic routing, b... more Open Access publishing has nothing to do with internet based technology or fiber-optic routing, but with old-school technologies from the 1980s. Photocopier are used to distribute printed journals and typewriters are the preferred choice in generating new knowledge. The mail room of the university library is the central hub for incoming and outgoing papers while a personal talk with the supervisor is the only choice for evaluating the quality. Modern technology like the OPAC digital catalog should be used very carefully and can not replace classical card catalog systems in the basement.
Ever wonder why the APC charge in a quality journal is so high? Perhaps you doesn't know, what ki... more Ever wonder why the APC charge in a quality journal is so high? Perhaps you doesn't know, what kind of underlying technology is used in producing the content. It is mostly outdated technology which is 20 years old and this results into huge amount of manual work until the paper can be published in pdf format. The idea of delaying important inventions is not a new idea, it was done by neoluddism movement many times in the history before. The general idea is, that technology doesn't break known workflows from the printed journal.
In computeranimation a motion graph is a learned model which describes potential transitions betw... more In computeranimation a motion graph is a learned model which describes potential transitions between states. It is a lowlevel constraint for specifying legal moves. The main reason for using it is because it helps to reduce the problem space. If it is clear, that the planning horizon only certain actions are allowed, it makes it easier for searching in the graph. Implementing a motion graph in sourcecode can be done with RRT planners, cognitive architectures, case-based reasoning and forward simulation. In the second part of the paper, some tips are given how to handle large software projects under the assumption that the best length of a single C++ file is limited to 140 lines of code.
A semi-autonomous robot control system works with trajectory tracking. The motion controller is f... more A semi-autonomous robot control system works with trajectory tracking. The motion controller is following the human-demonstration in realtime. The question is: how exactly? One possible answer is to store the rawdata in a file like the opening book of a chess-engine and access them with a query language. The state-action space is reduced to a feature vector which is similar to semantic datacompression. Neural networks can help to store the episodic memory in a machine readable form.
Forth is not known in mainstream computing, because the reverse polish notation is difficult to l... more Forth is not known in mainstream computing, because the reverse polish notation is difficult to learn. To overcome the problem, a Forth language simulator is described which makes it easier for newbies to play around with a multistack-machine. The idea is to reduce the performance down to 1 instruction per second so that the user can observe in a singlestep mode what his program is doing. The question which remains open is how to program in Forth itself. The development of stack based algorithm is outside the scope of this paper, here is only a simulator given, which is able to execute existing Forth code.
Learning from demonstration is an interactive method for developing a motion controller. The idea... more Learning from demonstration is an interactive method for developing a motion controller. The idea is, that a human-operator is driving a RC-car by hand, a motion capture system is tracking the position and a pid-controller is driving RC-car2 to the same position. Following a trajectory can done with a behavior tree (so called pid-controller), a neural network or a RRT-like planner. The following paper is a short introduction into the subject.
A preprint server which can be filled by everyone is only the first part of a document triage. Fo... more A preprint server which can be filled by everyone is only the first part of a document triage. For evaluating the content a "lab wiki" is needed, which is edited collaboratively. Without a dedicated convergent component , the workflow can't be called scientific. It remains an alchemy trial, which is equal to personal experience and told stories about publishing a paper.
Writing an academic paper can be done with LaTeX, MS-Word or Lyx. A recent development is the usa... more Writing an academic paper can be done with LaTeX, MS-Word or Lyx. A recent development is the usage of Google Docs, which is a combination of an online pdf-hoster and a word-processing tool. The problematic aspect of Google Docs is, that it can be used as part of social networks like Google Plus for distributed academic content creation. That means, that in theory the 500 Million users of Google Plus can act as a writing collective for producing high-quality science without any costs. It is unclear, if this thread has already started or not.
I don't think we should put Google and Facebook in the same basket. Facebook is part of much bigg... more I don't think we should put Google and Facebook in the same basket. Facebook is part of much bigger alliance, which tries to destroy Google.
An online encyclopedia is working with the same principles like an adventure game. The player has... more An online encyclopedia is working with the same principles like an adventure game. The player has to interact with the system and a score determines, if he was successful. The only difference is, that the feedback in an online encyclopedia is delayed. After uploading an edit, it won't happen anything. Filling the time gap is possible with the help of an abstract game description. With this book, new users will learn how to master the challenge.
Uploads