A view-based neurocomputational system for relational map-making and navigation in visual environments

Artificial navigation systems stand to benefit greatly from learning maps of visual environments, but traditional map-making techniques are inadequate in several respects. This paper describes an adaptive, view-based, relational map-making system for navigating within a 3D environment defined by a spatially distributed set of visual landmarks. Inspired by an analogy to learning aspect graphs of 3D objects, the system comprises two neurocomputational architectures that emulate cognitive mapping in the rat hippocampus. The first architecture performs unsupervised place learning by combining the “What” with the “Where”, namely through conjunctions of landmark identity, pose, and egocentric gaze direction within a local, restricted sensory view of the environment. The second associatively learns action consequences by incorporating the “When”, namely through conjunctions of learned places and coarsely coded robot motions. Together, these networks form a map reminiscent of a partially observable Markov decision process, and consequently provide an ideal neural substrate for prediction, environment recognition, route planning, and exploration. Preliminary results from real-time implementations on a mobile robot called MAVIN (the Mobile Adaptive VIsual Navigator) demonstrate the potential for these capabilities.     

Real-time trajectory generation in multi-RCCL

This article describes the design of the trajectory generator for a robot programming system called Multi-RCCL, which is a package of “C” routines for doing real-time manipulator control in a UNIX environment. RCCL has been used successfully in developing robot control applications in numerous research and industry facilities over the last several years. One of its strongest features is the ability to integrate real-time sensor control into the manipulator task specification. RCCL primitives supply the trajectory generator with target points for motions in joint or Cartesian coordinates. Other primitives allow the code developer to specify on-line functions that can modify the target points, or possibly cancel motion requests, in response to various sensor or control inputs. The design requirements of the trajectory generator are that it be able to integrate these on-line modifications into the overall robot motion and provide a smooth path between adjacent motions even when sensor inputs make the future trajectory uncertain. © 1993 John Wiley & Sons, Inc.     

Compound behaviors in pheromone robotics

We are pursuing techniques for coordinating the actions of large numbers of small-scale robots to achieve useful large-scale results in surveillance, reconnaissance, hazard detection, and path finding. Using the biologically inspired notion of “virtual pheromone” messaging, we describe how many coordinated activities can be accomplished without centralized control. By virtue of this simple messaging scheme, a robot swarm can become a distributed computing mesh embedded within the environment, while simultaneously acting as a physical embodiment of the user interface. We further describe a set of logical primitives for controlling the flow of virtual pheromone messages throughout the robot swarm. These enable the design of complex group behaviors mediated by messages exchanged between neighboring robots.     

The safety-bag expert system in the electronic railway interlocking system elektra

The real-time expert system “Safety-Bag” is an integral part of the electronic railway interlocking system ELEKTRA. The interlockinh system contains a logic channel, which processes commands, and a safety channel, which checks the commands according to safety rules. Commands which are only allowed under certain circumstances are carried out only if the instructions generated by processing the command in the logic channel are checked and committed by the safety channel. To minimize the possibility of common errors in both channels, different programming paradigms are used. The logic channel is implemented in a procedural programming language and the Safety-Bag is implemented in PAMELA, a rule-based expert system language suitable for real-time applications. The rule-oriented programming paradigm is very well suited to the Safety-Bag because the safety requirements themselves are represented by rules. The Safety-Bag runs on a standard ALCATEL process control computer.     

A robot-task conformance index for the design of robotized cells

A robot-task conformance index is introduced. It applies to any manipulator and a broad scope of tasks. Bounded in the [0, 1] range, this dimensionless index is defined within or outside singularities, and remains invariant against changes in units for translations or rotations. It is based on the simultaneous diagonalization of quadratic forms accounting for the “capabilities” of the robot and the “constraints” of the task. The structural and joint properties of the robot are formulated as a repeatability ellipsoid. A sufficient task ellipsoid is construed from the task's domain of constraints. Following the simultaneous diagonalization of both robot and task ellipsoids, minimum containing and maximum contained ellipsoids are deduced. Appropriate volume ratios of these four ellipsoids define the conformance index. Calculations performed in the diagonalized space reduce the computational burden.     

Intelligent robotic assembly

It is normal when programming a robotic manipulator to provide the end effector's orientation and position at the pick up and drop off locations. Additional sensory information and intelligence is needed, however, to detect the presence of a part as well as its location if the assembly site cannot be controlled precisely by employing expensive jigs or fixtures. This paper investigates, for this purpose, the application of solely an inexpensive laser sensor mounted unobtrusively to the end effector of a CRS robot having customized hardware and open software. Data from the sensor is converted into a single “Feature Value Vector” to recognize a part and accurately determine its location by using a neural network and back propagation training. The procedure's viability is tested by assembling a set of tightly meshing gears under poor ambient lighting.     

Teacher feedback to scaffold and refine demonstrated motion primitives on a mobile robot

Task demonstration is an effective technique for developing robot motion control policies. As tasks become more complex, however, demonstration can become more difficult. In this work, we introduce an algorithm that uses corrective human feedback to build a policy able to perform a novel task, by combining simpler policies learned from demonstration. While some demonstration-based learning approaches do adapt policies with execution experience, few provide corrections within low-level motion control domains or to enable the linking of multiple of demonstrated policies. Here we introduce Feedback for Policy Scaffolding (FPS) as an algorithm that first evaluates and corrects the execution of motion primitive policies learned from demonstration. The algorithm next corrects and enables the execution of a more complex task constructed from these primitives. Key advantages of building a policy from demonstrated primitives is the potential for primitive policy reuse within multiple complex policies and the faster development of these policies, in addition to the development of complex policies for which full demonstration is difficult. Policy reuse under our algorithm is assisted by human teacher feedback, which also contributes to the improvement of policy performance. Within a simulated robot motion control domain we validate that, using FPS, a policy for a novel task is successfully built from motion primitives learned from demonstration. We show feedback to both aid and enable policy development, improving policy performance in success, speed and efficiency.     

Performance evaluation of dynamically linearized and kinematically redundant planar manipulators

The objective of this research is to investigate the feasibility of a theoretical development which could lead to advanced manipulator concepts offering fast dynamic and control features. Specifically, the application of “dynamic linearization” to a kinematically redundant planar manipulator is studied. In comparing such a manipulator with a standard non-linear, non-redundant design it is shown that in addition to the expected merits in linearity, dexterity, and computational simplicity, savings in actuator energy consumption can also be realized using a relative simple trajectory optimization scheme.     

Neural network modelling of a compliant device to enhance and simplify compliant force control

A method is offered that uses an artificial neural network (ANN) to simplify the problem of mapping a robot tool position in space when using compliant force control. For this case, the compliant device (CD) is a moulded rubber block situated between the manipulator end flange and the tool. A force sensor is mounted between the tool and the CD. It is necessary to be able to relate the recorded contact forces and torques (hereafter referred to as “forces”) to the changes in translations and orientations (hereafter referred to as “positions”) of the tool relative to the manipulator. Usually, a complex Newtonian model would be required to achieve this. For this paper, to overcome the difficulties of accurately modelling the non-linear and highly coupled characteristics of the CD, back propagating ANNs have been trained to relate the forces to positions for real data sets. The finished ANN is to be used as part of a computer model of the overall manipulator system. This paper assesses the ability of the ANN to model the CD. In particular, three issues are addressed: first, the comparison of the ANN with a conventional spring type model; second, the verification of the ANN with true data sets; third, the need to filter the ANN output to avoid problems from outliers.     

Recognizing image “style” and activities in video using local features and naive Bayes

The goal of this paper is to offer a framework for classification of images and video according to their “type”, or “style”––a problem which is hard to define, but easy to illustrate; for example, identifying an artist by the style of his/her painting, or determining the activity in a video sequence. The paper offers a simple classification paradigm based on local properties of spatial or spatio-temporal blocks. The learning and classification are based on the naive Bayes classifier. A few experimental results are presented.     

A survey of socially interactive robots

This paper reviews “socially interactive robots”: robots for which social human–robot interaction is important. We begin by discussing the context for socially interactive robots, emphasizing the relationship to other research fields and the different forms of “social robots”. We then present a taxonomy of design methods and system components used to build socially interactive robots. Finally, we describe the impact of these robots on humans and discuss open issues. An expanded version of this paper, which contains a survey and taxonomy of current applications, is available as a technical report [T. Fong, I. Nourbakhsh, K. Dautenhahn, A survey of socially interactive robots: concepts, design and applications, Technical Report No. CMU-RI-TR-02-29, Robotics Institute, Carnegie Mellon University, 2002].     

Interval logics and their decision procedures : Part II: a real-time interval logic

In a companion paper, we presented an interval logic, and showed that it is elementarily decidable. In this paper we extend the logic to allow reasoning about real-time properties of concurrent systems; we call this logic real-time future interval logic (RTFIL). We model time by the real numbers, and allow our syntax to state the bounds on the duration of an interval. RTFIL possesses the “real-time interpolation property,” which appears to be the natural quantitative counterpart of invariance under finite stuttering. As the main result of this paper, we show that RTFIL is decidable; the decision algorithm is slightly more expensive than for the untimed logic. Our decidability proof is based on the reduction of the satisfiability problem for the logic to the emptiness problem for timed Büchi automata. The latter problem was shown decidable by Alur and Dill in a landmark paper, in which this real-time extension of ω-automata was introduced. Finally, we consider an extension of the logic that allows intervals to be constructed by means of “real-time offsets”, and show that even this simple extension renders the logic highly undecidable.     

Learning to Recognize and Grasp Objects

We apply techniques of computer vision and neural network learning to get a versatile robot manipulator. All work conducted follows the principle of autonomous learning from visual demonstration. The user must demonstrate the relevant objects, situations, and/or actions, and the robot vision system must learn from those. For approaching and grasping technical objects three principal tasks have to be done—calibrating the camera-robot coordination, detecting the desired object in the images, and choosing a stable grasping pose. These procedures are based on (nonlinear) functions, which are not known a priori and therefore have to be learned. We uniformly approximate the necessary functions by networks of gaussian basis functions (GBF networks). By modifying the number of basis functions and/or the size of the gaussian support the quality of the function approximation changes. The appropriate configuration is learned in the training phase and applied during the operation phase. All experiments are carried out in real world applications using an industrial articulation robot manipulator and the computer vision system KHOROS.     

Learning to Recognize and Grasp Objects

We apply techniques of computer vision and neural network learning to get a versatile robot manipulator. All work conducted follows the principle of autonomous learning from visual demonstration. The user must demonstra te the relevant objects, situations, and/or actions, and the robot vision system must learn from those. For approaching and grasping technical objects three principal tasks have to be done—calibrating the camera-robot coordination, detecting the desired object in the images, and choosing a stable grasping pose. These procedures are based on (nonlinear) functions, which are not known a priori and therefore have to be learned. We uniformly approximate the necessary functions by networks of gaussian basis functions (GBF networks). By modifying the number of basis functions and/or the size of the gaussian support the quality of the function approximation changes. The appropriate configuration is learned in the training phase and applied during the operation phase. All experiments are carried out in real world applications using an industrial articulation robot manipulator and the computer vision system KHOROS.     

Robotic clicker training

In this paper, we want to propose the idea that some techniques used for animal training might be helpful for solving human–robot interaction problems in the context of entertainment robotics. We present a model for teaching complex actions to an animal-like autonomous robot based on “clicker training”, a method used efficiently by professional trainers for animals of different species. After describing our implementation of clicker training on an enhanced version of AIBO, Sony’s four-legged robot, we argue that this new method can be a promising technique for teaching unusual behavior and sequences of actions to a pet robot.