Abstract — Robots that work alongside us in our homes and workplaces could extend the time an elderly person can live at home, provide physical assistance to a worker on an assembly line, or help with household chores. In order to assist us in these ways, robots will need to successfully perform manipulation tasks within human environments. Human environments present special challenges for robot manipulation since they are complex, dynamic, uncontrolled, and difficult to perceive reliably. In this paper we present a behavior-based control system that enables a humanoid robot, Domo, to help a person place objects on a shelf. Domo is able to physically locate the shelf, socially cue a person to hand it an object, grasp the object that has been handed to it, transfer the object to the hand that is closest to the shelf, and place the object on the shelf. We use this behavior-based control system to illustrate three themes that characterize our approach to manipulation in human environments. The first theme, cooperative manipulation, refers to the advantages that can be gained by having the robot work with a person to cooperatively perform manipulation tasks. The second theme, task relevant features, emphasizes the benefits of carefully selecting the aspects of the world that are to be perceived and acted upon during a manipulation task. The third theme, let the body do the thinking, encompasses several ways in which a robot can use its body to simplify manipulation tasks. 1 Fig. 1. The humanoid robot Domo used in this paper. I.

o the requirements or the egre1!1 P 1 '1 r P l 1 r 1 1

The architecture of control system for mobile robots is proposed in this paper. This architecture is based on hybrid approach using neural networks for classification of images and organization of associative memory as well as semantic networks for natural language processing and organization of memory and achievment of goals.

These controllers include smooth pursuit visual tracking, inverse kinematic reaching, and operation space control of the arm [77]. This layer also provides TCP/IP interprocess communication among the Linux cluster’s 1Gb LAN. We use the Yarp software package developed by Metta and Fitzpatrick [91]. We implemented a custom python-Yarp interface, allowing us to dynamically define and transmit data structures between processes at rates up to 100hz. Additionally, two FireWire framegrabbers provide synchronized image pairs to the cluster. Finally, all image and sensory data are timestamped using the hardware clock from the CANbus PCI card. This ensures synchronization of the data up to the transmit time of the 1Gb LAN. 4.7.4 Behavior Layer The behavior layer implements the robot’s visual processing, learning, and task behaviors. These algorithms are run within our behavior-based architecture named Slate. 4.8 Slate: A Behavior Based Architecture We have developed a behavior based architecture named Slate. What is meant by a robot architecture? According to Mataric [90], An architecture provides a principled way of organizing a control system. However, in addition to providing structure, it imposes constraints on the way the control problem can be solved. Following Mataric, Arkin [4] notes the common aspects of behavior-based architectures: • emphasis on the importance of coupling sensing and action tightly • avoidance of representational symbolic knowledge • decomposition into contextually meaningful units Roboticists have developed many flavors of behavior based architectures. We refer to Arkin for a review [4]. Loosely stated, Slate is a lightweight architecture for organizing perception and control. It is implemented as a programming abstraction in Python that allows one to easily define many small computational threads. These threads can run at parameterized 66 slate arbitrator thread threa s d fs slate module thread thread slate thread s a s proprioception yarp communication slate scheduler process module thread

This paper describes the current state of RatSLAM, a Simultaneous Localisation and Mapping (SLAM) system based on models of the rodent hippocampus. RatSLAM uses a competitive attractor network to fuse visual and odometry information. Energy packets in the network represent pose hypotheses, which are updated by odometry and can be enhanced or inhibited by visual input. This paper shows the effectiveness of the system in real robot tests in unmodified indoor environments using a learning vision system. Results are shown for two test environments; a large corridor loop and the complete floor of an office building. 1