Search

Games have been used as excellent testbeds for research on artificial intelligence and computational intelligence for their diversity and complexity. In this paper, we present AI World Cup, a set of AI competitions based on the game of soccer. We provide an introduction to the three challenges that concern a robot soccer match using both value-based and image-based state representations. AI Soccer runs the robot soccer match by participants managing each team of five two-wheeled robots. AI Commentator and AI Reporter observe the AI Soccer match and output real-time commentary and a summary article, respectively. Also, we introduce the AI World Cup platform along with rationale behind notable design choices. The official international AI World Cups held in 2018 and 2019, and the AI Masters competition held in 2019 as a part of the World Cyber Games are briefly discussed. Technical aspects of the strategies developed by participants are also discussed.

In human-robot interaction (HRI), classification is one of the most important problems, and it is essential particularly when the robot recognizes the surroundings and chooses a reaction based on a certain situation. Each interaction is different since new people appear or the environment changes, and the robot should be able to adapt to different situations during a brief interaction. Thus, it is imperative that the classification is performed incrementally in real time. In this sense, we propose an online incremental classification resonance network (OICRN) that enables incremental class learning in multi-class classification with high performance online. In OICRN, a scale-preserving projection process is introduced to use the raw input vectors online without a normalization process in advance. The integrated network of the convolutional neural network (CNN) for feature extraction and the OICRN for classification is applied to a robotic system that learns human identities through HRIs. To demonstrate the effectiveness of our network, experiments are carried out on benchmark data sets and on a humanoid robot, Mybot, developed in the Robot Intelligence Technology Laboratory, KAIST.

Robots are expected to perform smart services and to undertake various troublesome or difficult tasks in the place of humans. Since these human-scale tasks consist of a temporal sequence of events, robots need episodic memory to store and retrieve the sequences to perform the tasks autonomously in similar situations. As episodic memory, in this paper we propose a novel Deep adaptive resonance theory (ART) neural model and apply it to the task performance of the humanoid robot, Mybot, developed in the Robot Intelligence Technology Laboratory at KAIST. Deep ART has a deep structure to learn events, episodes, and even more like daily episodes. Moreover, it can retrieve the correct episode from partial input cues robustly. To demonstrate the effectiveness and applicability of the proposed Deep ART, experiments are conducted with the humanoid robot, Mybot, for performing the three tasks of arranging toys, making cereal, and disposing of garbage.

As the Korean demilitarized zone(DMZ) has unique geological features and terrains, conventional wheel-based unmanned ground vehicles(UGVs) are hard to perform the landmine removal tasks. Compared to other types of robot platforms, multi-legged robots, which climb the inclined way and overcome the complex obstacles with lower contact area on the ground, have several advantages to perform landmine detection tasks in DMZ effectively. Thus, to exploit such benefits, in this paper, we propose the design schemes and gait methods of multi-legged spider robots. Especially, this paper presents the design and motion control of multi-legged spider robots with four and six legs. In detail, we implemented the design and motion control of multi-legged spider robots suitable for landmine detection in the DMZ by applying the gait algorithms on Arduino and Dynamixel servomotors. Our proposed algorithm and system design demonstrate that the proposed system design architecture enables operational flexibilities while walking in harsh conditions.

Automated task planning for robots faces great challenges in that the sequences of events needed for a particular task are mostly required to be hard-coded. This can be a cumbersome process, especially, when the user wants a robot to learn a large number of similar tasks with different objects that are semantically related. We propose a novel approach of user preference-based integrated multi-memory model (pMM-ART). This approach focuses on exploiting a semantic hierarchy of objects alongside an episodic memory for enhancing the behavior of an autonomous agent. We analyze the functioning principle of the proposed model by teaching it a few distinct domestic tasks and observe that it is able to carry out a large number of similar tasks based on the semantic similarities between learned objects. We also demonstrate, via experiments using Mybot, our ability to reach those goals that are not possible without the integration of semantic knowledge with episodic memory.

For social human robot collaboration, robots need to effectively remember human experiences and manage emotional experiences as well as repetitive experiences. To implement these functions, the hierarchical emotional episodic memory, using deep adaptive resonance theory network, is proposed in this paper. The proposed memory not only learns emotional experiences, but also has the ability to anticipate future emotional situations. Two parameter modulation processes, delayed consolidation and instant update, are provided. These make emotional experiences reinforce faster, remain for longer, and become more stable and sensitive to analogous experiences. Simulation analysis is conducted to verify the proposed memory, and an experiment is carried out in a kitchen environment to demonstrate social human robot collaboration.

The crux of the realization of task intelligence for robots is to design the memory module for storing temporal event sequences of tasks, the mechanism of thought for reasoning, and motion planning methodology for execution, among others. In this paper, task intelligence is realized using episodic memory, neural model-based mechanism of thought, and an online motion planning algorithm. Robots are taught either by demonstration or symbolic description. A behavior appropriate to the current situation is selected by the developmental episodic memory-based mechanism of thought, while a proper task is retrieved from Deep adaptive resonance theory (ART). The behaviors are executed safely and quickly with the proposed motion planning algorithm. The effectiveness and applicability of task intelligence are demonstrated through experiments with the humanoid robot, Mybot, developed in the Robot Intelligence Technology Laboratory at KAIST.

In order to perform various tasks using a robot in a real environment, it is necessary to learn the tasks based on recognition, to be able to derive a task sequence suitable for the situation, and to be able to generate a behavior adaptively. To deal with this issue, this paper proposes a system for realizing task intelligence having a memory module motivated by human episodic memory, and a task planning module to resolve the current situation. In addition, this paper proposes a technique that can modify demonstrated trajectories according to current robot states and recognized target positions in order to perform the determined task sequence, as well as a technique that can generate the modified trajectory without collisions with surrounding obstacles. The effectiveness and applicability of the task intelligence are demonstrated through experiments with Mybot, a humanoid robot developed in the Robot Intelligence Technology Laboratory at KAIST.

Intelligent robots such as social robots and home service robots that are being developed nowadays are required to have the ability to naturally communicate with users in natural language. Beyond simple data retrieval and simple dialogue level, we propose a new Human-Robot Interaction (HRI) system that enables a robot to understand and reason the environment around a user and present information about them in natural language, whenever the user asks a question to the robot. For its intelligent HRI, based on Dynamic Memory Networks (DMN), a neural network for Visual Question Answering (VQA), we propose a new full sentence VQA network model called Full-Sentence Highway Memory Network (FSHMN). For its robot platform, a three DOF robotic head was used which has a neck with three motors and a tablet PC head. To verify the feasibility of the proposed system, an experiment is performed in which a user and a robot interact with each other in a way of question answering in a customized kitchen environment. Through the experiment, we not only demonstrated the effectiveness of applying deep learning to HRI applications in real environments but also presented a new insight into HRI.

Nowadays the application of robots are emerging in areas of modern life. It is expected we will be living in a new era in which robots such as socially interactive robots will make an important effect on our daily lives. Considering emotions play a critical role in human social communication, emotional episodes are necessary for human-robot social interactions. In this regard, we propose a framework that can form a social relationship between human and robot using an emotional episodic memory. Proposed framework enables personalized social interactions with each user by identifying the user and retrieving the matching episode in the memory. The interaction is not fixed, the emotional episodic memory is developmental through additional experiences with the user. The proposed framework is applied to an interactive humanoid robot platform, named Mybot to verify the effectiveness. As demonstration scenarios, photo shooting, and user identification and robot’s emotional reactions are used.

The automated home referred to as Smart Home is expected to offer fully customized services to its residents, reducing the amount of home labor, thus improving human beings' welfare. Service robots and Internet of Things (IoT) play the key roles in the development of Smart Home. The service provision with these two main components in a Smart Home environment requires: 1) learning and reasoning algorithms and 2) the integration of robot and IoT systems. Conventional computational intelligence-based learning and reasoning algorithms do not successfully manage dynamic changes in the Smart Home data, and the simple integrations fail to fully draw the synergies from the collaboration of the two systems. To tackle these limitations, we propose: 1) a stabilized memory network with a feedback mechanism which can learn user behaviors in an incremental manner and 2) a robot-IoT service provision framework for a Smart Home which utilizes the proposed memory architecture as a learning and reasoning module and exploits synergies between the robot and IoT systems. We conduct a set of comprehensive experiments under various conditions to verify the performance of the proposed memory architecture and the service provision framework and analyze the experiment results., Accepted (Early Access)

Task Intelligence is the capacity of a robot to learn, reason and execute specific behaviours based on its environment. In this paper, the Task Intelligence problem formulated by the Robot Intelligence and Technology Laboratory at KAIST is researched further: specifically the proposed contribution is a brand new perceptual pipeline in which the recognition, detection, segmentation and grasping of objects is achieved assuming no prior knowledge of the environments arrangement nor the objects appearance. A Convolutional Neural Net (CNN) is used to detect, recognize and semantically label those objects that need to be interacted with. 3D point clouds, corresponding to the objects model, are extracted after several segmentation procedures and registered over time. Dimensional and positional information of the object is acquired. Additional grasping pose data is calculated. All of the collected knowledge is parsed so that the Task Intelligence system is able to deal with previously unknown objects in dynamic environments. This system is formed by an Episodic Memory (Deep-ART), an action sequence generator (FF-planner) and a trajectory warping module for pre-learnt behaviours. The proposed approach has been tested using the Webots simulator.

To prepare for the anticipated age of human–robot symbiosis, robots should be able to interact and cooperate with humans effectively by understanding the meaning and intention of human behavior. In this paper, we define human intention as “desired behavior of the human using objects.” To infer the defined human intention, a robot should learn the object affordance along with a behavior hierarchy structure. Thus, in this paper, we propose a behavior hierarchy-based affordance network (BHAN) and a behavior hierarchy-based affordance map (BHAM) to represent the object affordance, behavior hierarchy structure, and object hierarchy structure, simultaneously. Autonomous and interactive BHAN/BHAM learning algorithms are also proposed to make a robot develop the BHAN and BHAM by itself, as well as by interacting with a human. Based on the newly developed BHANs and BHAM, a robot could infer the human intention from information observed in context and from human behavior. The effectiveness of the proposed method was demonstrated through experiments on human–robot interaction with building blocks using a simulated differential wheel robot and a real human-sized humanoid robot.

Research on developing human-like gaze control has been carried out to enhance human-robot interaction. From the viewpoint of a large consistency of human gaze, conventional research had focused on predicting where humans usually pay attention to. However, gaze control is a cognitive process that can even produce different scanpaths from the same visual information. In this paper, an evolutionary fuzzy integral-based gaze control algorithm with preference is proposed. It produces various scanpaths according to the preference of human gaze. The proposed gaze control algorithm evaluates each pixel point with fuzzy measures and fuzzy integral, and produces a scanpath through repeated selections considering memory and bio-inspired processes. The produced scanpath is transformed into a fixation map and compared with a scanpath obtained from a human subject by the earth mover’s distance. Based on the comparison, quantum-inspired evolutionary algorithm gradually develops preference of human gaze to produce a scanpath similar to the human scanpath. The effectiveness of the proposed algorithm is demonstrated by comparing a human scanpath with a scanpath produced from the algorithm using the developed preference. The applicability of the proposed algorithm is also demonstrated by applying the developed preference to gaze control for learning from demonstration.

Locating a fire inside of a structure that is not in the direct field of view of the robot has been researched for intelligent firefighting robots. By classifying fire, smoke, and their thermal reflections, firefighting robots can assess local conditions, decide a proper heading, and autonomously navigate toward a fire. Long-wavelength infrared camera images were used to capture the scene due to the camera’s ability to image through zero visibility smoke. This paper analyzes motion and statistical texture features acquired from thermal images to discover the suitable features for accurate classification. Bayesian classifier is implemented to probabilistically classify multiple classes, and a multiobjective genetic algorithm optimization is performed to investigate the appropriate combination of the features that have the lowest errors and the highest performance. The distributions of multiple feature combinations that have 6.70% or less error were analyzed and the best solution for the classification of fire and smoke was identified.

To perform a home service task through cooperation with a human in a real environment, a robot needs to deal with the environmental changes and accordingly plan appropriate behavior sequence. For this purpose, in this paper, we propose an adaptive task planner which is based on memory and reasoning. A robot perceives user behaviors and objects using an RGB-depth and thermal sensor. The robot stores a temporal sequence of behaviors for performing a task in its episodic memory that is realized by a sequence to sequence network. When the user command is given, the episodic memory is used to retrieve the behavior sequence to carry out the command. On the other hand, when the robot perceives user behaviors, the robot postpones its behavior till his/her behavior is stopped. Once stopped, the episodic memory retrieves the behavior sequence to conduct a task that the user has intended. A task scheduler schedules the behavior sequence from the memory and sends it to an internal simulator. The internal simulator confirms the behavior sequence to be executable and then if executable, it sends the next executable behavior to the execution module. If a behavior fails in the internal simulation test, fast forward planner generates an alternative behavior sequence to resolve the failed behavior problem. The effectiveness and applicability of the proposed planner is demonstrated by a wheel-based humanoid robot.

To use a service robot in a public place, the robot should be able to adapt its behaviors and learn from experiences with people. However, the existing behavior selection algorithms have difficulties in learning deeply or in creating a new behavior. To solve these problems, we implement the human-like thought process, i.e. developmental episodic memory-based mechanism of thought (DEM-MoT), in behavior selection of social robots. In DEM-MoT, a behavior is selected among pre-defined behaviors by an adaptive resonance theory based on the 2-norm Euclidean distance. If the expected reward of the selected behavior is too low, a new behavior is created by randomly arranging a set of primitive behaviors. To show the effectiveness of the proposed behavior selection method, experiments were carried out in a simulated environment.

This paper proposes a novel approach for the autonomous 3D model generation of unknown objects. A humanoid robot (or any setup with two manipulators) holds the object to model in one hand, views it from different perspectives and registers the depth information using a RGB-D sensor. The occlusions due to limited movement of the manipulator and the gripper itself covering the object are avoided by switching the object from one hand to the other. This allows for additional viewpoints leading to the registration of more depth information of the object. The contributions of this paper are as follows: 1. A humanoid robot that manipulates objects and obtains depth information 2. Tracing the hand movements with the robots head to be able to see the object at every moment 3. Filtering the point clouds to remove parts of the robot from them 4. Utilizing the Normal Iterative Closest Point algorithm (depth points, surface normals and curvature information) to register point clouds over time. This method will be applied to those pointclouds that include the robots gripper for optimal convergence; the resultant transform is then applied to those point clouds that describe only the segmented object 5. Changing the object from one hand to another 6. Merging the resulting object’s partial point clouds from both the left and right hands 7. Generating a mesh of the object based on the triangulation of final points of the object’s surface. No prior knowledge of the objects is necessary. No human intervention nor external help (i.e visual markers, turntables ...) is required either.

We developed an interactive humanoid robotic platform with a real-time face learning algorithm for user identification and an emotional episodic memory to combine emotional experiences with users so that the robot can differentiate its reactions to users according to the emotional history. In this video, it is demonstrated how a robot can develop a social relationship with humans through the face identification and emotional interaction.

This paper presents Fast Batch Informed Trees (Fast-BIT∗), a modification to the asymptotically optimal path planner Batch Informed Trees (BIT∗). Fast-BIT∗ modifies the underlying heuristic that dictates the expansion and processing of vertex and edge queues. BIT∗ uses heuristics to guide the search of implicit Random Geometric Graphs (RGGs) to generate an explicit solutions, while minimizing highly computational tasks such as collision checking. Fast-BIT∗ builds on BIT∗ by biasing the search heuristic towards the goal, in a solution analogous to depth-first search, finding an initial solution of the implicit RGG at a faster rate, at the cost of decreasing initial optimality. Fast-BIT∗ requires additional procedures to reset expansion variables of affected areas in the graph, ensuring the bias is not lasting in the graph expansion. An earlier initial solution leads to a faster initial upper bound for use in informed graph pruning, allowing convergence of path cost to begin earlier in the planning procedure. We show that Fast-BIT∗ finds a first solution faster than BIT∗ as well as the commonly used RRT-Connect and similar methods, along with a faster overall convergence rate. This paper shows various random-world test examples, showing the benefits of similar algorithms, along with a robot path planning simulation.

The end-to-end approach presented in this paper deals with the recognition, detection, segmentation and grasping of objects, assuming no prior knowledge of the environment nor objects. The proposed pipeline is as follows: 1) Usage of a trained Convolutional Neural Net (CNN) that recognizes up to 80 different classes of objects in real time and generates bounding boxes around them. 2) An algorithm to derive in parallel the pointclouds of said regions of interest (ROI). 3) Eight different segmentation methods to remove background data and noise from the pointclouds and obtain a precise result of the semantically segmented objects. 4) Registration of the object's pointclouds over time to generate the best possible model. 5) Utilization of an algorithm to detect an array of grasping positions and orientations based mainly on the geometry of the object's model. 6) Implementation of the system on the humanoid robot MyBot, developed in the RIT Lab at KAIST. 7) An algorithm to find the bounding box of the object's model in 3D to then create a collision object and add it to the octomap. The collision checking between robot's hand and the object is removed to allow grasping using the MoveIt libraries. 8) Selection of the best grasping pose for a certain object, plus execution of the grasping movement. 9) Retrieval of the object and moving it to a desired final position.

This paper proposes a resource-oriented, decentralized auction algorithm (RODAA) for multirobot task allocation considering multiple resources of the robots and limited robot communication range. The resources that this paper focuses on are the expendable supplies that a robot consumes and recharges while performing tasks, such as energy. In the proposed algorithm, each robot generates its cost for the task in a probabilistic manner considering multiple paths that visit none or different combinations of refill stations for performing the task based on the robot's residual resources. For robust and time-efficient task allocation with limited robot communication range in a dynamic network, a multihop-based auction algorithm is proposed. This paper also introduces a solar panel cleaning mission as a new application for multirobot systems and the proposed algorithm is implemented in the simulation of the mission. The simulation results demonstrate that the proposed algorithm is capable of completing the panel cleaning mission faster than other auction-based task allocation algorithms and has lower overall resource consumption.

During the last few decades, as a part of effort to enhance natural human robot interaction (HRI), considerable research has been carried out to develop human-like gaze control. However, most studies did not consider hardware implementation, real-time processing, and the real environment, factors that should be taken into account to achieve natural HRI. This paper proposes a fuzzy integral-based gaze control algorithm, operating in real-time and the real environment, for a robotic head. We formulate the gaze control as a multicriteria decision making problem and devise seven human gaze-inspired criteria. Partial evaluations of all candidate gaze directions are carried out with respect to the seven criteria defined from perceived visual, auditory, and internal inputs, and fuzzy measures are assigned to a power set of the criteria to reflect the user defined preference. A fuzzy integral of the partial evaluations with respect to the fuzzy measures is employed to make global evaluations of all candidate gaze directions. The global evaluation values are adjusted by applying inhibition of return and are compared with the global evaluation values of the previous gaze directions to decide the final gaze direction. The effectiveness of the proposed algorithm is demonstrated with a robotic head, developed in the Robot Intelligence Technology Laboratory at Korea Advanced Institute of Science and Technology, through three interaction scenarios and three comparison scenarios with another algorithm.

This paper proposes a novel online multiobjective evolutionary approach for the navigation of humanoid robots. In the proposed approach, the humanoid robot navigation problem is decomposed into a series of small multiobjective optimization problems (MOPs) with corresponding local information. Using multiobjective evolutionary algorithms (MOEAs), the MOPs can be successively solved while the robot is walking. In addition, to achieve significant reductions in the processing time of the MOEAs for online implementation while maintaining robustness and scalability, a novel homogeneous parallel computing method is devised for the MOEAs. Multiobjective particle swarm optimization with preference-based sort (MOPSO-PS) is employed as the MOEA to reflect the user-defined preference for each objective during navigation. The effectiveness of the proposed online approach is demonstrated through well-known benchmark problems and a robot simulator. In both the simulation and the experiment, a humanoid robot successfully navigates to the goal, satisfying the preferences for various objectives, with local information in an environment without a global map.

A real-time probabilistic classification method was developed for identifying fire, smoke, their thermal reflections, and other objects in infrared images. This algorithm was formulated for use on a robot that will autonomously locate fires inside of a structure where the fire is outside the robot field of view. Thermal images were used to extract features due to the fact that long wavelength infrared is capable of imaging through zero visibility environments. For an autonomous navigation under fire environments, robots need to be able to differentiate between desired characteristics, such as fire and smoke, and those that may lead the robot in the incorrect direction, such as thermal reflections and other hot objects. The probabilistic classification method in this paper provides a robust, real-time algorithm that uses thermal images to classify fire and smoke with high accuracy. The algorithm is based on four statistical texture features identified through this work to characterize and classify the candidates. Based on classification of candidates from large-scale test data, the classification performance error was measured to be 6.8% based on validation using the test dataset not included in the original training dataset. In addition, the precision, recall, F-measure, and G-measure were 93.5–99.9% for classifying fire and smoke using the test dataset.

The objective of multirobot task allocation (MRTA) is to assign tasks to robots to minimize the overall cost of task performance. This paper proposes a decentralized MRTA approach considering the robots' residual expendable resources and their limited communication ranges. The proposed approach consists of two algorithms, namely, resource-aware cost generation (RCG) and ad hoc network-based task allocation (ANTA). The RCG algorithm allows each robot to generate a credible cost of task performance considering its residual resources in task planning. The ANTA algorithm constructs the minimal spanning tree network among the robots and determines the robot with the lowest cost for the task through multihop communication in a decentralized manner. The advantages of the proposed approach are minimization of unnecessary task performance cost caused by resource shortage of robots during task execution and the use of an ad hoc network among the robots to allow more robots to participate in the task allocation process. The effectiveness of the proposed approach is demonstrated through computer simulations for multirobot foraging as a test problem.

One of the major functions of intelligent robots such as social or home service robots is to interact with users in natural language. Moving on from simple conversation or retrieval of data stored in computer memory, we present a new Human-Robot Interaction (HRI) system which can understand and reason over environment around the user and provide information about it in a natural language. For its intelligent interaction, we integrated Dynamic Memory Networks (DMN), a deep learning network for Visual Question Answering (VQA). For its hardware, we built a robotic head platform with a tablet PC and a 3 DOF neck. Through an experiment where the user and the robot had question answering interaction in our customized environment and in real time, the feasibility our proposed system was validated, and the effectiveness of deep learning application in real world as well as a new insight on human robot interaction was demonstrated.

This paper presents a fire heading estimation for solving the autonomous navigation problem of a firefighting robot in smoke-filled indoor fire environment. In smoke-filled fire environments, firefighters and firefighting robots experience difficulty maintaining direction while finding the fire source. To solve this, the statistical texture features in thermal images were analyzed and fused by using Bayesian estimation to compute the vertical and horizontal fire heading. For its validation, a large-scaled test-bed was built with a hallway and two rooms, with one of the rooms having a real size fire generating dense and dark smoke. The proposed method probabilistically computed the fire-heading toward the entrance of the hallway then guided the robot to the room with the actual fire, all while navigating in a smoke-filled situation. The experimental results have demonstrated the effectiveness of this method in indoor fire environments.

By the number of people aged 60 or over and people with disabilities growing, homecare mobile robot draws increasing attention. However, there are challenges of autonomous navigation for homecare robot such as frequent changes of environment, obstacles and goal position. In this paper, we focus on verifying potential of neural network-based autonomous navigation for homecare mobile. And we compare recurrent neural network with multilayer perceptron in the navigation of an autonomous mobile robot. The result suggested that the recurrent neural network can do better robot navigation because of its capability to handle the temporal dependency of a data sequence. Also, it shows that neural network-based navigation can be a good alternative since it has decent generalization ability for new environment, obstacles and goals.

Since the industrial revolution, technological evolution has been the defining aspect of societal progress. Among all the technologies that have emerged over the years, information technology is the one that truly revolutionized the modern lifestyle. To take information technology beyond information mining and communication, intelligence technology (InT) is required to manipulate the information and knowledge with human-like intelligence. However, transforming InT into the next technological wave requires a medium to actualize InT with all its potential. Therefore, in this paper, we propose an intelligence super agent (iSA) as an autonomous entity equipped with all categories of intelligence, i.e. cognitive, social, behavioral, ambient, collective, and genetic intelligences. To realize the iSA, we also present a modular architecture, named intelligence operating architecture (iOA), capable of implementing all aspects of InT along with a biologically-inspired thought process. As a novel implementation platform for the iSA, we present user-centered development of robots enacted through agent based modularization (UDREAM). The effectiveness of UDREAM is shown through some application-specific experiments with individual robots. Although the current iSA is for a limited number of services with a standalone robot, it can be extended to a more complex iSA.

In this paper, an evolutionary-optimized central pattern generator (CPG) considering equality constraints is proposed for stable modifiable bipedal walking. The proposed CPG generates the position trajectories of the swing foot and the center of pelvis in the Cartesian coordinate system at single and double support phases. The significance of the proposed CPG is that it can change the sagittal and lateral step lengths just before the beginning of each single support phase while maintaining the desired values of single and double support times, which are set in the beginning of bipedal walking. To deal with environmental perturbations, the sensory feedbacks in the CPG are designed using the force sensing resistors such that the bipedal robot can maintain its balance. For the optimized parameters of the CPG, a two-phase evolutionary programming is employed. The effectiveness of the method is demonstrated by computer simulation with the Webots model of a small-sized humanoid robot, HSR-IX, and the experiment with HSR-IX developed in the RIT Laboratory, KAIST, Daejeon, Korea.

This paper proposes a method for stable bipedal walking with a vertical center-of-mass (COM) motion by an evolutionary optimized central pattern generator (CPG). To generate a walking pattern for a bipedal robot, a modifiable walking pattern generator (MWPG) is employed, which extends a conventional 3-D linear inverted pendulum model (3-D LIPM) to allow a zero-moment-point variation by the closed-form functions. By using the MWPG, the robot is able to modify the walking pattern in real time while walking. For the vertical COM motion of the 3-D LIPM, the vertical COM trajectory is generated by the CPG. The disturbance caused by the vertical COM motion is compensated by utilizing the sensory feedback in the CPG. To obtain the desired output signals from the CPG, the CPG is optimized by the two-phase evolutionary programming (TPEP), which is suitable to solve the constrained optimization problems. By combining the MWPG with the CPG, stable bipedal walking with a larger stride is obtained. The validity of the proposed method is verified through real experiments for the small-sized bipedal robot, HanSaRam-IX.

Localization is a crucial ability for autonomous robots and landmark-based localization can be effectively used because it enables localization with only landmark information. For localization, the omnidirectional vision system is efficiently used for robots to obtain information of the surrounding environment, but it is expensive and has some distortion. In this sense, the fish-eye lens vision system can be an alternative. Compared to the omnidirectional vision system, however, it obtains less landmark information at a time so that it needs a localization algorithm using less landmark information. To solve this problem, this paper proposes a novel landmark-based particle localization algorithm for the global localization problem called relocation. It can localize the pose of the robot using only two landmarks. In this algorithm, information on bearing angle and distance of landmarks is used to calculate a possible area of the location of the robot and then particles, each of which represents a pose of the robot, are randomly distributed in the calculated area. The pose of the robot is identified by selecting a particle with the highest importance weight among the distributed particles. Computer simulations and experiments demonstrate the effectiveness of the proposed algorithm.

This article proposes a framework for competency adjustment and workload balancing in multirobot task allocation. Competency represents the ability of a robot to execute a task in terms of quality and cost, and workload balancing denotes distribution of workload among robots. This framework considers the quality and cost of a robot for a task and adjusts them in accordance with environmental changes. For workload balancing, the framework utilizes the concept of subsidy to encourage participation from the less active members of the robot team. The proposed framework is implemented in a simulated cleaning mission. Simulation results demonstrate that this framework can adjust the competency in accordance with environmental changes and distribute workload among robots in a balanced manner.

The ability to think has been one of the most fascinating features of robots in science fiction since the introduction of the word 'robot' by Karel Capek in his Czech science fiction play, known as 'Rossum's Universal Robots' in English [1]. Since Capek's first depiction in 1921, science fictionists generally portray robots as intelligent humanoid machines that are subservient to humans. Android is another term used for such robots in modern linguistics. However, the development of robots with the ability to think has largely been a fantasy for robot scientists and engineers. Nevertheless, the evolution of robots has come a long way since the first industrial robot 'Unimate' in 1954. Robots have transformed from large automated manufacturing facilities to applications in our homes and even into our pockets as software robots in our PDAs and smart phones [2], [3]. This paper classifies robot intelligence into six categories: cognitive intelligence, social intelligence, behavioral intelligence, ambient intelligence, collective intelligence and genetic intelligence. This classification categorizes the intelligence of robots based on the different aspects of awareness and the ability to act deliberately as a result of such awareness.

Bipedal robots in previous research work were unable to independently modify the elements of a walking pattern on uneven terrain without any extra footstep for adjusting the center of mass (COM) motion. To solve this problem, this paper extends the modifiable walking pattern generator (MWPG). The MWPG can independently modify the elements of the walking pattern without any extra footstep on flat terrain only. The extended MWPG can be applied on uneven terrain. In the extended MWPG, a 3-D command state is defined to generate the walking pattern on uneven terrain. Instead of using the constant COM height, the vertical COM trajectory is generated to satisfy the foot height of the swing leg. Also, an additional trajectory, generated by the cubic spline interpolation, is supplied to the vertical foot trajectory of the MWPG. The proposed MWPG is implemented on the small-sized bipedal robot, HanSaRam-IX (HSR-IX) and the effectiveness of the proposed MWPG is demonstrated through the experiment.

This paper proposes a novel scheme of integrating episodic memory into semantic memory based task planner. Task planners have taken an important role in AI research along with semantic memory to better perform tasks for robots. Episodic memory memorizes and retrieves temporal sequence of situated behaviors by which temporal relationship between behaviors can be defined. None of any research, however, has implemented it into their work for task planning. By introducing episodic memory into task planner, the temporal causal relationship between situated behaviors, which are stored in semantic memory, is taken into consideration. The integrated architecture proves its effectiveness by notably reducing the number of nodes traversed in finding solutions. Robots can reduce time complexity in solving given problems by retrieving previous memories. Deep Adaptive Resonance Theory (Deep-ART) neural model and cogency-based hierarchical behavior planner are used for the episodic memory and the task planner, respectively. Cogency-based hierarchical behavior planner proves its capability of solving given problems in experiment with humanoid robot Mybot, and Deep-ART is augmented to the planner and tested in simulations. Therefore, the contribution of this approach lies on developing a framework which takes advantage of implementing episodic memory and planner in one place.

This paper proposes a novel hierarchical behavior planner with a multi-layered confabulation based behavior selection structure for robots to perform tasks. The proposed planner integrates a STRIPS based behavior selection approach and cogent confabulation approach. The STRIPS based behavior selection approach is a goal tree search that induces goal-oriented sequences of behaviors, while the cogent confabulation approach is based on conditional probabilities between input symbols and target behaviors, aims to model human thinking mechanism. Our planner is applied with a set of behaviors defined in a multi-layered structure to show that it can plan a hierarchical sequences of behaviors to perform given tasks. The effectiveness and applicability of the proposed scheme is demonstrated through the experiments with the robot Mybot, developed in the Robot Intelligence Technology Lab. at KAIST.

Intelligent behaviors generally mean actions showing their objectives and proper sequences. For robot, to complete a given task properly, an intelligent computational model is necessary. Recurrent Neural Network (RNN) is one of the plausible computational models because the RNN can learn from previous experiences and memorize those experiences represented by inner state within the RNN. There are other computational models like hidden Markov model (HMM) and Support Vector Machine, but they are absent of continuity and inner state. In this paper, we tested several intelligent capabilities of the RNN, especially for memorization and generalization even under kidnapped situations, by simulating mobile robot in the experiments.

Robots can learn knowledge by observing demonstration of humans. As tutees, robots need to not only observe human behaviors, but also make proper feedbacks for human tutors because learning is an interactive process in which information is delivered in bidirectional ways between humans and robots. Gaze is an adequate method for robots to provide human tutors with feedbacks that robots are concentrating on current learning because gaze directly represents where they are paying attention to. This paper proposes a gaze control algorithm with a state machine in learning from demonstration. A human tutor shows demonstration in front of a robot tutee, and the robot tutee observes the demonstration for learning. The robot tutee perceives external environment through its camera, recognizes a human and objects, and figures out a state at which the robot tutee is situated. Then, the robot tutee gazes at proper targets that are predefined by the state machine. The human tutor also adjusts the demonstration to make learning more effectively according to the robot tutee’s feedbacks. The effectiveness of the proposed method is demonstrated through the experiments with a robotic head with 17 degrees of freedom, developed in the RIT Lab., KAIST.

This paper proposes a novel algorithm, named dual multiobjective quantum-inspired evolutionary (DMQEA) algorithm augmented fuzzy cognitive map (FCM). DMQEA was developed to help users select preferable solutions out of the non-dominated ones and has been proven to be an effective way compared to other multi-objective optimization methods, such as MQEA, MQEA-PS, etc. DMQEA, in this paper, has been coupled with decision supporting tool, fuzzy cognitive maps (FCMs) to support selecting best models which can reflect users' preferences. Even though the attempts with single optimization such as genetic algorithms (GAs) or particle swarm optimization (PSO) have been frequently carried out, there have been only few attempt to incorporate FCM with multicriteria decision making algorithm, especially to reflect user's preference. This study aims to integrate DMQEA with FCM to build models according to user's preference. In robotics field, the interaction with human operators is an important issue and DMQEA-FCM can aid robots in their decision making process in the context of the interaction.

This paper proposes a new concept of the genetic robot which has its own robot genome, in which each chromosome consists of many genes that contribute to defining the robotpsilas personality. The large number of genes also allows for a highly complex system, however it becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the robotpsilas personality while manually initializing values for the individual genes. To overcome this difficulty, this paper proposes an evolutionary algorithm for a genetic robotpsilas personality (EAGRP). EAGRP evolves a gene pool that customizes the robotpsilas genome so that it closely matches a simplified set of features desired by the user. It does this using several new techniques. It acts on a 2 dimensional individual upon which a new masking method, the Eliza-Meme scheme, is used to derive a plausible individual given the restricted preference settings desired by the user. The proposed crossover method allows reproduction for the 2-dimensional genome. Finally, the evaluation procedure for individuals is carried out in a virtual environment using tailored perception scenarios.

As the robot soccer system becomes stabilized, it has been used as an educational platform with which various topics on mobile robotics can be taught. As one of key topics in the education of mobile robotics is computational intelligence-based navigation, this paper proposes a multiobjective population-based incremental learning (MOPBIL) algorithm to obtain the fuzzy path planner for optimal path to the ball, minimizing three objectives such as elapsed time, heading direction and posture angle errors in a robot soccer system. MOPBIL employs the probabilistic mechanism, which generates new population using probability vectors. As the probability vectors are updated by referring to nondominated solutions, population converges to Pareto-optimal solution set. Simulation and experiment results show the effectiveness of the proposed MOPBIL from the viewpoint of the proximity to the Pareto-optimal set, size of the dominated space, coverage of two sets and diversity metric. By implementing each of the solutions into the educational platform, it can be educated how multi-objective optimization is realized in the real-world problem.

An artificial creature has its own genome in which each chromosome consists of many genes that contribute to defining its personality. The large number of genes allows a highly complex system. It becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the artificial creature's personality, if gene values are manually assigned for the individual genome. To solve this problem, this paper proposes a multi-objective evolutionary generation process for artificial creatures' specific personalities (MOEGPP), where the dimension of the personality model is defined as that of optimization objectives. Key components of MOEGPP are as follows: i) the complement of (1 - k) dominance; ii) the pruning method considering objective deviation for all genomes;and iii) the inutation using biased normal distribution, to get a set of nondominated genomes with specific personalities according to the defined personality dimension. By proposed MOEGPP, nondominated genomes having specific personalities can be obtained and they are successfully tested by using an artificial creature, Rity, in the virtual 3D world created in a PC.

This paper proposes a novel vector field navigation method by using the velocity vector of a robot to avoid obstacles in dynamic environment. Shifting vector and virtual obstacle are proposed to derive the center of repulsive field which makes the robot move away from the obstacle: The shifting vector is derived from the velocity vector of the robot and the obstacle and the virtual obstacle is made by adding the shifting vector to the real obstacle. By considering the velocity vector of an obstacle with a shifting vector and a virtual obstacle, the robot can avoid not only stationary obstacles but also dynamic obstacles, where the fields are optimized by evolutionary programming. The performance of proposed method is demonstrated by simulations and experiments in robot soccer system.

This paper proposes a CSMA/CD-R (Carrier Sense Multiple Access/Collision Detection with Reservation) protocol, designed for wireless network mobile robots under a distributed robot system without any centralized mechanism. It employs stations to reserve a communication channel after communication collision for the wireless communication of the distributed robot system. The effectiveness and applicability of the proposed protocol are demonstrated by carrying out computer simulations and real experiments with the develped multi-robot system.

A software robot requires plausible external features for intimate interaction with humans. The various appearance designs effectively contribute to provide him/her with an opportunity to select a preferable robot among them. This paper proposes genetic representation for the appearance of software robot, which is inspired by homologous chromosomes. User selection scheme is based on the interactive genetic algorithm. As user evaluates the preference of external features of software robot, proposed algorithm converges to the most preferable appearance to user's choice. The effectiveness of the proposed scheme is demonstrated through experiments.

One of the key components of an ubiquitous robot (Ubibot) is the software robot (Sobot), which can communicate with embedded robot (Embot) and mobile robot (Mobot). Sobot is a virtual robot, which has the ability to move to any place or to connect to any device through a network without any spatial limitation. Embot has the capability to sense the surroundings and to interpret the context of the environment and can communicate with Mobot and Sobot. Mobot provides an integrated mobile services. To incorporate Sobot, Embot, and Mobot reliably as an Ubibot, a middle layer is needed to arbitrate different protocols among them. This paper focuses on incorporating Sobot and Mobot to overcome physical limitations of Sobot for physical behaviors in real situations. To implement the incorporation of them, the basic concept and structure of the middle layer are proposed. The effectiveness of the middle layer for Sobot and Mobot is demonstrated through real experiments.

A robot is expected to carry out a task autonomously with its own knowledge system. Using the knowledge system, the robot can recognize current situation and recall a proper sequence for performing an appropriate task in that situation. To build such knowledge system, the robot learns the knowledge from user demonstrations as if a child learns through interactions with parents and teachers. User demonstration is captured by an RGBD camera embedded the robot. The robot needs to segment each execution from continuous RGB-D streams. In this paper, each execution is composed of an object and an action performed on the object. The sequence of executions, or the procedure, should be stored in the robot's memory for the the robot to retrieve and execute the procedure in a similar situation later. Such a procedural memory is developed based on an adaptive resonance system. Using the procedural memory learned, the robot can perform the full sequences of tasks with only partial information given on executions. The effectiveness of the proposed scheme is demonstrated for four tasks through computer simulations.

Human-like gaze control of robots is essential for natural human robot interaction (HRI), and thus it is necessary to identify the gaze control factors that affect human gaze. In this paper, the gaze control factors are derived inductively from the observation of human scanpaths. Human scanpaths are measured from movie clips and are transformed into vector-based representation that contains spatio-temporal information. The transformed scanpaths are compared to find the most common scanpath. The selected scanpath is analyzed to identify the gaze control factors. The derived factors are anticipation, knowledge, and attractions in a static image such as salient regions, humans, objects, and center of visual inputs.