The Mobile Robot Rhino

Abstract

s RHINO was the University of Bonn's entry in the 1994 AAAI Robot Competition and Exhibition. RHINO is a mobile robot designed for indoor navi-gation and manipulation tasks. The general scien-tific goal of the RHINO project is the development and the analysis of autonomous and complex learning systems. This article briefly describes the major components of the RHINO control software as they were exhibited at the competition. It also sketches the basic philosophy of the RHINO archi-tecture and discusses some of the lessons that we learned during the competition. R HINO, shown in figure 1, is a B21 mobile robot platform manufactured by Real-World Interface. It is equipped with 24 sonar proximity sensors, a dual-color camera system mounted on a pan-tilt unit, and 2 on-board I486 computers. Sonar information is obtained at a rate of 1.3 hertz (Hz), and cam-era images are processed at a rate of 0.7 Hz. RHINO communicates with external computers (two Sun SPARCSTATIONs) by a tetherless Ether-net link. The RHINO project is generally concerned with the design of autonomous and complex learning systems (Thrun 1994). The 1994 AAAI Robot Competition and Exhibition, sponsored by the American Association for Artificial Intelligence (AAAI), ended an initial six-month period of software design. Key fea-tures of RHINO's control software, as exhibited at the competition, are as follows: Autonomy: RHINO operates completely autonomously. It has been operated repeated-ly for durations as long as one hour in popu-lated office environments without human intervention. Learning: To increase the flexibility of the software, learning mechanisms support the adaptation of the robot to its sensors and the environment. For example, neural network learning is employed to interpret sonar mea-surements. Real-time operation: To act continuously in real time, any-time solutions (Dean and Boddy 1988) are employed wherever possible. Any-time algorithms are able to make deci-sions regardless of the time spent for compu-tation. The more time that is available, how-ever, the better the results are. Reactive control and deliberation: RHINO's navigation system integrates a fast, reactive on-board obstacle-avoidance routine with knowledge-and computation-intense map building and planning algorithms. RHINO's software consists of a dozen differ-ent modules. The interface modules (a base-sonar sensor interface, a camera interface, and a speech interface) control the basic commu-nication to and from the hardware compo-nents of the robot. On top of these, a fast obstacle-avoidance routine analyzes sonar measurements to avoid collisions with obsta-cles and walls at a speed as high as 90 cen-timeters a second. Global metric and topolog-ical maps are constructed on the fly using a neural network–based approach combined with a database of maps showing typical rooms, doors, and hallways. RHINO employs a dynamic programming planner to explore unknown terrain and navigate to arbitrary target locations. It locates itself by continu-ously analyzing sonar information. In addi-tion, a fast vision module segments images from two color cameras to find target objects and obstacles that block the path of the robot. RHINO's control flow is monitored by an integrated task planner and a central user interface. The integration of a dozen different soft-ware modules, which all exhibit different tim-ing and response characteristics, requires a flexible scheme for the flow and