The development of papa goose robot type started in 2001 with the very start of robotics activities at Jacobs University. From 2002 on, it was assisted by mother goose, a smaller tracked robot. The nicknames papa- and mother-goose are by the way derived from the fact that they are intended to cooperate with a set of small autonomous robots called the ducklings. The papa goose as well as the mother goose robots are meanwhile replaced by the rugbot design.
Papa goose robots were used by the Jacobs University Rescue Robot team in several competition including the RoboCup worldchampionship 2002 in Fukuoka Japan, the RoboCup worldchampionship 2003 in Padua Italy, the RoboCup American Open 2004 in New Orleans USA, and the RoboCup worldchampionship 2004 in Lisbon Portugal.
Like most Jacobs University robots, papa goose robots are based on complete in-house designs, ranging from the mechanics over sensors and actuators to the software level. This allows to optimize the designs for the particular tasks like rescue operations. For this purpose the CubeSystem, a collection of hardware and software components for fast robot prototyping, is used.
Locomotion
Papa-goose is based on a differential six-wheel drive. It is rather large with a footprint of 450mm × 400mm to allow for a full-fledged PC in addition to its sensors and drive components as well as a significant payload. Its on-board batteries allow for roughly 1 hour of intensive use in difficult terrain and up to 5 hours of operation with sparse locomotion.
Its locomotion system is a differential drive with six wheels that behaves similar to a tracked drive. The three wheels of each side are driven via belts and a motor-unit connected to the axis of the rear, respectively front wheel on the left, respectively right side of the robot. The first implementation in 2001 was based on a 1:115 spur gear. From 2002 on, the spur gears were replaced by a 66:1 planetary gear-boxes. The motors have 90 W power each and they are equipped with HP quadrature encoder with 500 pulses per channel.
Papa goose is carrying in addition to its CubeSystem an onboard PC. To get an easy start, we chose a standard desktop configuration for this purpose. The on-board PC is hence based on a rather large ATX motherboard with a Pentium III processor running at 1 GHz. The disadvantage of the size is compensated by the 6 PCI slots on the motherboard that allow a significant flexibility for adding cards that can be used to interface various sensors. Already for the most basic configuration of the robot, 3 PCI slots are needed, namely one for a 4-port USB 2.0 extension for cameras, one for a frame grabber for a thermacam, and one for IEEE 802.11 RF-ethernet.
Sensors and Software
The robots are equipped with four USB cameras with a resolution of 1280×960 pixels. At the RoboCup worldchampionship 2003 in Padua Italy, the Jacobs University rescue robot team introduced a very special sensor for victim detection into the recue competition, namely a thermal camera. The Flir P60 thermacam used on a rugbot as an uncooled, high resolution Focal Plane Array (FPA). Its 320 x 240 elements provide temperature information in a range of -40oC to 120oC with 0.08oC resolution. The color to temperature map can be changed such that the related image highlights only spots with human body temperature. During the competition, the sensor turned out to be very useful as it helps to spot victims under conditions where normal visual feedback completely fails. The standard lens of the camera provides a field of view of 24 deg x18 deg. This field of view was found by our operator to be a bit narrow, especially when compared to the wider opening angles of the normal cameras used on the robots. Hence special optics with field of views of 45 deg x 34 deg were used later on. The main disadvantage of thermacams is cost.
Other sensors on the robots are microphones that can also be used to detect victims in rescue scenarios. As obstacle sensors low-cost laserscanners from Hokuyo Automatic, the PB9-11, are used. It covers 162 degrees in 91 steps up to a depth of 4m. This sensor is the main tool for gathering obstacle data. The bases are in addition equipped with several one-dimensional obstacle sensor, namely coarse range ultrasound sensors with a long range of up to 10 m and a wide scan angle of 60 deg, high precision ultrasound sensors with a medium range of up to 7 m and a narrow scan angle of 10 deg, and active infrared sensors with a short range of up to 0.7 m and narrow scan angle of 10 deg. These sensors are ideal for simple control behaviors like wall-following to autonomously negotiate long corridors. Last but not least, papa goose is equipped with bumpers in form of safety switch tapes. These bumpers are are based on special profile rubber tubes with a highly sensitive switch matrix. The rubber tubes have some shock-absorption capabilities and are thus suited as ultimate safety measure to stop the robot in case that it crashes despite the other sensors into an obstacle.
To estimate the absolute orientation of a papa goose robot, two digital compasses are used. The first one is based on the Philips KMZ51 IC. It has an I2C interface and it is directly connected to the CubeSystem. The second compass is from Honeywell. Its RS232 interface is serviced by the onboard PC. The motors of the robots are equipped with high resolution quadrature encoders from HP. The software modules of the CubeSystem not only use this data for control, but also for odometry and dead-reckoning to estimate the robot’s pose. In doing so, the data from the compass is used for a leaky update of the orientation estimation via odometry. By this, the performance of dead-reckoning gets significantly improved. This can be explained by the fact that the odometry based estimation of orientation severely suffers from cumulative error and hence significantly drifts. The absolute orientation measurements of the compass compensate this drift. Using papa goose robots, the Jacobs University rescue robot team was at the RoboCup worldchampionship 2003 in Padua Italy the very first team that managed to autonomously generate map in the unstructured environment of a rescue competition.