Research article

A search and rescue robot with tele-operated tether docking system

Carlos Marques, Joao Cristovao and Paulo Alvito
IdMind-Engenharia de Sistemas, Lda, Lisboa, Portugal, and

Pedro Lima, Joao Frazao, Isabel Ribeiro and Rodrigo Ventura

Institute for Systems and Robotics, Instituto Superior Tecnico, Lisboa, Portugal
Abstract Purpose To describe a robot designed and built to operate in outdoor environments hostile to the human presence, such as debris resulting from the collapse of built structures, and targeted to the tele-operated detection of potential survivors using a set of specic sensors whose information is transmitted to a remote human operator. Design/methodology/approach RAPOSAs mechanical structure is composed of a main body and a front body, whose locomotion is supported on tracked wheels, allowing motion even when the robot is upside down. The front body has variable tilting capabilities, providing means to overcome edges higher than the robot main body (e.g. when climbing a stair) and is also useful to grab the lower ground when only the main body has ground contact. This front body has one thermal camera and two webcameras installed. Additional sensors include gas, temperature and humidity sensors, web cams, light diodes, microphone and loudspeaker. The robot uses wireless communications, with an option for tethered operation. Findings The robot was tested in several scenarios of the Fire Fighters school. In this particular exercise, the robot reduced the inspection time down to 25 percent of the time that specialized reghters teams would take to nish the exercise. This was due to the fact that the reghters need to stabilize the environment in order to reduce live threats. In this case, as in many other similar situations, not only the robot provides a faster inspection method, but also a much safer one. Originality/value The tether carries both power and communications, with an access point on its end. Docking and undocking the robot to the tether is accomplished remotely by the operator with the help of a camera located inside the robot, and represents the most innovative feature of RAPOSA. Keywords Search and rescue, Robotics, Fire services Paper type Research paper

1. Introduction
Search and rescue in emergency scenarios arising from natural and man-made disasters is one important application of mobile robots. After an earthquake or the collapse of a built structure, and facing a scenario of large destruction, the response time to search and locate trapped survivors is crucial as it is known that after 48 h of the disaster the probability of survival is low. The human intervention of urban search and rescue (USAR) teams, including USAR dogs, has to be done cautiously so as to protect the rescue workers from further collapses. Debris may be so cluttered that prevent the close human access to the victims. Also, potential risk of further landslide requires the propping of the structures before human intervention. Rescue preparation operations may be time consuming, and a fast action to locate survivors and to
The current issue and full text archive of this journal is available at www.emeraldinsight.com/0143-991X.htm

take them human voices, light and/or water is a crucial factor for life. Therefore, there is the clear need for search and rescue robots that are small, cheap and light, and that can be released immediately after a disaster in which the conditions are too dangerous and too cluttered for people and dogs to begin searching for victims. The most well-known work on USAR robots in the USA has been carried out by Casper and Murphy (2003), namely on the usage of several tele-operated robots for real search and rescue missions, in cooperation with professional human teams, including the participation in the rescue operations of the World Trade Center (WTC), after the September 11 attacks. The National Institute of Standards and Technology has also developed the USAR Performance Metrics and Test Arena (Jacoff et al., 2003), a real scenario which emulates several real-world situations faced by human teams after an earthquake, which has been widely used worldwide, e.g. in Europe, at the Intelligent Systems for Emergencies and Civil Defense in Rome, Italy, and which has been serving as the
This work was partially supported by Fundacao para a Ciencia e a Tecnologia (ISR/IST plurianual funding) through the POS_Conhecimento Program that includes FEDER funds, and by the Portuguese Agency for the Innovation.

Industrial Robot: An International Journal 34/4 (2007) 332 338 q Emerald Group Publishing Limited [ISSN 0143-991X] [DOI 10.1108/01439910710749663]

332

A search and rescue robot with tele-operated tether docking system Carlos Marques et al.

Industrial Robot: An International Journal Volume 34 Number 4 2007 332 338

testbed for the RoboCup Rescue initiative (Tadokoro et al., 2000), which joins together annually dozens of teams in a search and rescue robotic competition. This strong worldwide interest in search and rescue robots research and development has attracted several companies that have developed USAR commercial platforms, such as iRobot (Packbot, a very robust, light and mobile tele-operated robot, used in the WTC operations but recently re-targeted towards military operations), RoboProbe Technologies (bomb disarming and inspection robots), Inuktun (tele-operated robots also used in the WTC operations), Foster-Miller (demining and bomb-disarming robots, some of them appropriate for USAR operations, also used in the WTC operations), or the South Korean Domy and Co., whose robots can be tele-operated using wireless communications and provide remote audio interaction with victims. One common feature of these platforms is that teleoperation is possible either using wireless communications or a tether, but not both. Accordingly, when the tether is used, the onboard batteries are useless, since it is not possible to switch from the tether- to the batteries-supplied power during normal operation, without changing the robot structure. The robot described in this work combines tether-supplied wireless communications, tether- and battery-supplied power, exchangeable during operation, to take advantage of the positive features of wireless and tethered solutions. This paper describes the robot RAPOSA, shown in Figure 1, designed and built to operate in outdoors environments hostile to the human presence, such as debris resulting from the collapse of built structures. The robot is targeted to the tele-operated detection of potential survivors using a set of specic sensors whose information is transmitted to a remote human operator. An innovative feature of our work is the use of wireless communications, with an option for tethered operation. The tether carries both power and communications, with a wireless transceiver on its end, and can also be used to suspend the robot inside a deep hole. Docking and undocking the robot to the tether is accomplished remotely by the operator with the help of a camera located inside the robot. The paper is organized as follows: Section 2 describes the mechanical structure of the robot and the design constraints imposed by the environments where the robot is supposed to operate. Sensing capabilities and the mixed tethered/wireless communication system, the most innovative feature of RAPOSA, are described in Section 3. Section 4 covers the Figure 1 Robot RAPOSA: external view

graphical user interface (GUI), a relevant part of this teleoperated robot. A description of several tests made so far in very realistic environments is made in Section 5. Section 6 closes the paper, drawing the main conclusions and listing topics of interesting future work in the robot.

2. Mechanical structure
The robot RAPOSA is targeted to be used in USAR operations, in particular in debris resulting from collapsed or unstable man-made structures. The project specication in all the aspects related with mechanical design, sensors, GUI and functional capabilities was dened in a close collaboration with the Lisbon Fire Fighters and Portuguese Civil Protection Departments. This allowed a categorization of the several types of scenarios to consider, the obstacles to overcome and also the scenarios where this robot should not operate, given their extreme complexity (e.g. underwater). The more important constraints consider that the robot should t in sewer pipes with a standard diameter of 40 cm used as a way to reach locations otherwise unaccessible in disaster scenarios and should be able to climb and descend stairs with steps of standard dimensions of 17 cm of height 23 cm of width (Figure 2). Those constraints determined the major components of the mechanical design, namely: . Two modules, a main body and a frontal body, whose relative vertical orientation with respect to the main body is adjustable. . Two-side tracked wheels to provide locomotion for both modules. The frontal body locomotion is coupled to that of the main body. . When the robot ips upside down, it continues its operation awlessly. This implies that the robot does not have a top or bottom part, and that it self-detects ipping and automatically exchanges the commands to the motors and ip the cameras images, as an example of adjustable autonomy. RAPOSA dimensions, meeting the above specications, are 17.5 cm tall, 37 cm wide and 75 cm long. Total weight is 27 Kg, with batteries included. The front body features two webcams, each with an associated light and a thermal camera. The two webcams assemblage provides a 308 horizontal pan that, associated with the front body ^ 908 tilt range, enables a large eld of view. Three 5Ah Li-Ion batteries were chosen to be on the front part of the main body (Figure 3), as it is crucial that the robot center of mass is located on its front, so that the robot falls to its head when climbing stairs, rather than the opposite (ipping or falling). On the front of the main body, there are Figure 2 The robot RAPOSA inside a sewer pipe and downstairs

333

A search and rescue robot with tele-operated tether docking system Carlos Marques et al.

Industrial Robot: An International Journal Volume 34 Number 4 2007 332 338

Figure 3 Robot RAPOSA: inside view

two wheels on each side. One is connected to the locomotion track on the main body and the other to the front body track. They are attached in such a way that the movement of the main body wheels is transmitted to the front body wheels, both rotating at the same speed, and the front body positions itself without interfering with the locomotion.

3. Sensing capabilities and communication

The temperature and humidity sensors provide a measurement of both relative humidity from 0 to 100 percent and temperature measurement from 240 to 1208C in a single sensor, using a digital interface. This sensor is located close to the gas sensor, in such a way that its measurements are used for humidity and temperature compensation for the gas sensors. Four different gas types can be detected by the on-board sensors: methane, propane, butane and other gases that indicate high explosive level, hydrogen sulde and carbon monoxide. Four webcams are installed on the robot: two on the front of the tilting arm, providing a exible eld of view to the remote operators, one on the front of the main body, and one on the robot rear. Besides, providing data for the perception of the nearby environment, the rear webcam also supports the docking operation of the communications and power cable. In dark environments, articial illumination is provided by low consumption light LEDs installed nearby the cameras, as shown in Figure 4. Figure 4 Front body with cameras (thermal and webcams), lights and IR sensors

In a disaster scenario, there is usually a considerable amount of dust in the air so, even with articial illumination, no distinguishable image at all can be retrieved using conventional cameras. A thermal camera, on the other hand, is sensitive to heat radiation, thus allowing the perception and detection of heat sources. This is very useful to help nding survivals under debris or dust. Figure 5, on the GUI, shows on top the images of the two webcams and, on the bottom, the thermal image of a potential survivor. Moreover, the thermal camera data can be used to estimate the temperature in a given zone, warning that a re may hide behind a hidden door or wall. A Raytheon Series 300 Digital (thermal) Camera with 18 mm lens is placed in the central location of the front arm. The camera is capable of detecting people up to 150-200 m, weights less than 1 Kg and provides a greyscale image at 30 frames per second. Real experiments of the robot tele-operation in realistic scenarios have shown that, particularly when overcoming large slopes (e.g. a stair), the perception of the distance to the terrain is of great help. For this purpose, a set of infra-red sensors was installed on the front body, facing down. The robot is able to climb 458 inclinations and, if it ips, the operator should be aware of that fact, as the image acquired by the camera(s) does, sometimes, provide an elusive idea of the correct robot orientation. Furthermore, the robot is allowed to operate turned upside down. Therefore, when ipping is detected by onboard sensors, the image is inverted and the meaning of joystick left-right commands is exchanged. Flipping requires frequent adjustments of roll and pitch estimates. Two analog tilt sensors were installed to measure and provide the operator with the knowledge of roll and pitch angles of the robot. Similar information for the front body is provided and displayed in the GUI as shown in Figure 5. For complete determination of the robot inclination, two orthogonally assembled accelerometers, measuring gravity, are required. The two-axis devices used have a limited range of about 70-758. This accelerometer-based solution computes, by double integration, the robot inclination with respect to the Earth surface, but it is affected by the robot accelerations. However, this effect can be minimized by low Figure 5 GUI operation console

334

A search and rescue robot with tele-operated tether docking system Carlos Marques et al.

Industrial Robot: An International Journal Volume 34 Number 4 2007 332 338

pass ltering the output of the sensor, using an embedded low pass lter with a cutoff frequency of 6 kHz. Concerning communications, existing robots, such as the iRobot Packbot or Domy and Co.s robots, have one of two main congurations: tethered or wireless. Both solutions have strong and weak points. The tethered solution provides better autonomy and ensured bandwidth. It can also be used to sustain or pull the robot. However, a cable may get stuck, broken, etc. thus limiting the robot mobility. The wireless solution, on the other hand, is less dependent on the terrain where the robot moves and the number of turns it has to make. Its autonomy depends solely on batteries. Nevertheless, wireless communications may prove very unreliable. Standard wireless LAN devices can reach a maximum of 50 m indoors, in good conditions. This is not the case on disaster scenarios where twisted metal, big piles of concrete, all kinds of obstacles, edges, electrical wires, etc. block and reect the signal, making it difcult to communicate at high data rates (or to communicate at all). The feedback received from other search and rescue teams advises the use of a tether. In many cases the electromagnetic noise is too high and wireless communications may not work at all. A cable, although being a dead weight, provides stable power and communication. Traditional solutions allow either congurations, but even if the same robot supports both of them, the change must be done at the setup stage, being a time consuming job. The solution proposed for this robot goes a step further, allowing the tether to be attached/detached whenever necessary in real time and while the mission is undergoing. To that purpose, a docking mechanism was installed on the back of the robot. The robot comes close to the cable, grabs it and attaches it, through a perpendicular lock. The robot can be operated with or without a cable and the switch can be made remotely in real time. The cable carries both power and communication. Power is supplied to the robot throught the electric contacts on the tether end, while communication is delivered by an wireless transceiver at that same end. The wireless communication follows the IEEE 802.11b (WiFi) standard. Either managed or ad hoc modes can be used, although the managed one is usually employed (the robot serves as an access point, where the cable end and/or the Operational Console are its clients). When the cable is connected to the robot, communication is assured by a shortrange wireless link between the robot and the cable end. Otherwise, a wireless link can still be established between these two points, with the advantage that connectivity only requires signal propagation between the robot and the tether end. Furthermore, in small scale operations, the robot can communicate directly, via a wireless link, with the Operational Console, thus avoiding the use of the cable. The current length of the cable is 30 m and is must be unwound manually as the robot moves and drags it. The cable enclosure was designed to be able to support the robot weight. This can be useful in real scenarios, e.g. to pull the robot out of a sealed pipe. The docking system is composed of two parts: the tether part, that is released on the ground, and the grabbing mechanism on the robot back. RAPOSA features an opening in the back, where a tether terminal enters and gets locked. The robot is equipped with a motorized locking mechanism, together with a rear camera (Figure 6). The operator moves the robot, using the visual feedback from the rear camera, to force the terminal at the tether end (Figure 7) entering the 335

Figure 6 Docking system, from the robot body side

locking mechanism. Then, the motorized locking mechanism closes, thus providing a solid grasp of the tether. The lock is strong enough to hold the robots weight, so that it can be lowered by the tether into a hole. If the tether is not required anymore, it is pulled off. As the robot moves, the tether releases itself from the robot. The cable is exible, but ends in a solid structure that has a pyramidal shape. This allows both unrestricted movement and a way to raise the bi-conical metal guide so that the robot can grab it. The pyramidal structure has a weight on the cable side and does not rotate easily, even if dropped in a nonhorizontal plane and thus the bi-conical metal guide maintains its orientation approximately, independently of the way the structure is dropped on the ground (Figure 8). The average distance from the ground to the end of the biconical metal guide was projected to be at the same height of the robot docking hole. Nevertheless, in the docking phase the vertical alignment can be cleverly done: if the frontal arm is pushed down, the main body rear comes closer to the ground. On the robot back, two sliding doors are able to release or grab the cable part. When the doors are closed the cable part is pulled to the robot inside due to its bi-conical metal guide. Meanwhile, electrical power spring contacts are pressed against two concentrically arranged rings on the back of the robot. The rings are made of conductive material. Each ring is connected to a voltage pole. Since, the spring contacts are also Figure 7 RAPOSA docking to its cable

A search and rescue robot with tele-operated tether docking system Carlos Marques et al.

Industrial Robot: An International Journal Volume 34 Number 4 2007 332 338

Figure 8 Details of the docking system cable end

concentrically arranged, the poles are never inverted. Before attaching and prior to detachment of the docking system the power cable must be turn off to prevent electrical glitches. The power contacts and bi-conical metal guide are separated from the pyramid body structure through a large spring, whose purpose is to avoid breaking the docking mechanism when it is dragged to unfavorable positions, normally when the robot starts climbing an obstacle. The need of assuring physical contact of four Ethernet terminals once the cable is locked, for communications purposes, is quite demanding. We avoided the problem by using a wireless transceiver at the end of the cable to communicate wirelessly with the robot, where two circular polarization antennas are located in the rear, nearby the cable. Although there is no physical contact, transmission is assured in the best possible conditions, since the distance between antennas is very small. The power transmission (DC voltage to power the robot) needs to have physical contact, however. It is fundamental to have a rear camera to aid the docking process. The camera was placed inside the robot, behind the insertion hole, aligned with it, so that it is useful even when the robot is upside down. When the docking mechanism is not attached and the operation scenario has no dust, the docking hole can be open and this camera used to have a view of the environment on the back of the robot. Besides, the Ethernet to wireless adaptor and corresponding antenna, the pyramid also features a small DC-DC 2A board power, the wireless bridge and four green LEDs. If the pyramid is dropped on a dark environment, this helps nding it again. The fact that the LEDs are also at an equal distance from the bi-conical guide helps precise maneuvering to grab the cable end.

state of the docking mechanism of the power/communication cable. The second and third are setup views for robot and sensor thresholds conguration. On these views the user can adjust the sensors and command rates, as well as to establish limits on the front body movement and to dene limits on the motors speeds (Figure 9). The human operator can also set the warning and danger values of the Roll and Yaw sensors and to turn on or off a specic sensor. The fourth view is an advanced debug view, where the operator can watch and control in real time the low level micro-controller data tables. More conguration settings can be included on this view if necessary. The commutation between views is possible selecting each of the corresponding tabs on the top of the interface. A game pad interface is used to control the robot on the eld. This is thought to be a better way to control the robot on the eld than the usual joystick. The game pad (Figure 10) is composed of two joysticks, one slider, a four key cursor and several buttons. One of the joysticks controls the robot motion, the other pans and tilts the front cameras. The slider is intended to control the front body position. The other buttons are used to select the active cameras, the LED lights intensity, to latch/unlatch the docking mechanism, to reverse the robot motion direction, as well as to enable and disable the motors. An agent-based software architecture, including different types of agents that can be combined both hierarchically and in a distributed manner, was used. The architecture supports information fusion between several sensors and the sharing of information between the agents by a blackboard and is geared towards the cooperation between robots. Agents are generically organized hierarchically. At the top of the hierarchy, the algorithms associated with the agents are likely to be planners (in this case replaced by the human operator), whilst at the bottom they are interfaces to control and sensing hardware. The planner agents are able to control the execution of the lower level agents to service highlevel goals. The latter can be distributed across several processors and/or robots. To offer platform independence, only the lowest level Figure 9 GUI rst setup view

4. User interface and software architecture

RAPOSA normal operation is to be controlled by an human operator through a GUI. The GUI is composed by four views. The rst view (Figure 5) is the Operation Console, used to drive the robot, in which the user has the camera (web and thermal) feedback images as well all the data from the robot sensors and the sensor and actuator commands, namely, on the top left, the pitch and roll angles, on the top right the arm position and the lights intensity, on the bottom right all the sensor values, as well as the network state, and on the bottom left the battery state, the motors velocity and the 336

A search and rescue robot with tele-operated tether docking system Carlos Marques et al.

Industrial Robot: An International Journal Volume 34 Number 4 2007 332 338

Figure 10 User interface game pad

Figure 12 RAPOSA gets inside a pipe, through a very narrow hole in the rubble

agents are specic to the hardware, and these have a consistent interface for communication with the planning agents that control their execution. The elements of the architecture are the agents, the blackboard, and the control/ communication ports (Frazao and Lima, 2004).

5. Tests in realistic scenarios

The robot was tested in several scenarios of the Lisbon Fire Fighters school, in March 2005. One of the scenarios consisted of a 40 m pipe hidden below a great amount of rubble, shown in Figure 11, which the robot traversed almost entirely (left side of Figures 2 and 12). At some points, large tires located inside the pipe could not be overcome by the robot, that even ipped when trying to step over one of them, thus enabling the test of the adjustable autonomy system. This worked quite well, and the combination of pitch and roll sensors with the front cameras was fundamental, since given the pipe radial symmetry it is hard to gure out whether the robot is in its natural position or not. These tests also allowed using the cable to pull the robot out of the pipe in Figure 11 RAPOSA moving in the Lisbon Fire Fighters school destruction scenario

case everything else fails. Figure 8 shows an exercise actually made with the robot, which mimics a standard exercise the Fire Fighters do with their search and rescue dogs, where a person gets inside a hole in the rubble, inside a pipe, and simulates he/she is trapped. The robot was teleoperated and the victim was visible in the infrared cam, as seen in the picture. Another scenario concerned the operation inside a twooor house (right side of Figure 2). The operator stayed outside and the robot was able to climb and descend stairs twice, as well as to undock and dock the power cable remotely, in a room with total absence of light. The robot did also traverse successfully a dark tunnel with a step at the end. The only minor problems encountered concerned wireless communications, both related to the antennas location on the robot body and interference with other wireless networks. In these exercises, the robot reduced the inspection time down to roughly 25 percent of the time that specialized re ghter teams would take to nish the exercise. This percentage is a rough estimate of the ratio robot operation time vs humans operation time, provided by the experienced re ghters. They usually arrive at the scene, prop up unstable rubble (an operation that takes a considerable time) and only afterwards get inside and attempt to nd victims. The same procedure applies to the dogs. The main issue is to make sure one does not miss human lives while looking for survivors that might not be there. With a robot, the operation can proceed without the overhead of propping up, and the human/dog team will only act if the robot nds survivors (and gives them some extra hope, a signicant boost to their life expectancy). Overall, the robot performed awlessly, and the Fire Fighters are willing to use it in real operations. In October 2005, RAPOSA was included on the Portuguese Search and Rescue team of the Catastrophes Intervention Department of Lisbon Fire Fighters, that participated in the international exercise Eurosot 2005, October 13-16, in Sicilia, Italy. In this simulated earthquake, RAPOSA was used to explore a pipeline system of a collapsed building. In May 2006, the robot represented Portugal in the ELROB2006 competition/demo, where it performed several operations in the urban scenario, including traversing grass elds, climbing 337

A search and rescue robot with tele-operated tether docking system Carlos Marques et al.

Industrial Robot: An International Journal Volume 34 Number 4 2007 332 338

ramps and stairs, opening doors by pushing and climbing up and down stairs. The remote docking mechanism was specially appreciated by the organizers of the event and other participants.

whole team but only, e.g. the leader of a robot formation. In the single-robot case, current work concerns autonomous docking using visual servoing techniques, and autonomous stairs climbing using appropriate motion patterns (e.g. zigzagging).

6. Conclusions and open issues

This paper described a robot designed and developed for USAR operations, namely its mechanical structure, sensing capabilities and communications, operator interface and software architecture. The robot uses wireless communications, with an option for tethered operation. The tether carries both power and communications, with a wireless transceiver on its end. Docking and undocking the robot to the tether is accomplished remotely by the operator with the help of a camera located inside the robot, and represents the most innovative feature of RAPOSA. The robot was tested in very realistic scenarios, and a rst prototype is expected to be used in real missions as soon as required. There are strong prospects to start its commercialization, after solving current minor problems with wireless communications and modications of the mechanical structure, following lessons learned during the preliminary tests. Prospective applications to building surveillance and road tunnel inspections are also under consideration currently. Future envisaged work concerns the widening of the adjustable autonomy capabilities, both to single-robots and to multi-robot teams, so as to free the operator from the most tedious work, and to help her/him locating victims and driving the robot(s) adequately. This may be especially interesting for robotic teams, where the operator would not need to drive the

References
Casper, J. and Murphy, R.R. (2003), Human-robot interactions during the robot-assisted urban search and rescue response at the world trade center, IEEE Trans. on Systems, Man and Cybernetics Part B, Vol. 33 No. 3, pp. 367-85. Frazao, J. and Lima, P. (2004), Agent-based software architecture for multi-robot teams, paper presented at the 5th IFAC/EURON Symposium on Intelligent Autonomous Vehicles IAV2004. Jacoff, A., Weiss, B. and Messina, E. (2003), Evolution of a performance metric for urban search and rescue robots, Proceedings of the 2003 Performance Metrics for Intelligent Systems (PerMIS) Workshop. Tadokoro, S., Kitano, H., Takahashi, T., Noda, I., Matsubara, H., Hinjoh, A., Koto, T., Takeuchi, I., Takahashi, H., Matsuno, F., Hatayama, M., Nobe, J. and Shimada, S. (2000), The Robocup-rescue project: a robotic approach to the disaster mitigation problem, paper presented at the IEEE International Conference on Robotics and Automation, pp. 4089-94.

Corresponding author
Pedro Lima can be contacted at: pal@isr.ist.utl.pt

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints

338