1. INTRODUCTION
Robotic aids, used widely in areas such as healthcare and home assistance, require cooperationbetween humans and robots. The proper controlcommands for such systems would be a combinationof autonomous commands from robots and commandsfrom humans, rather than either autonomouscommands or solely human commands [1-3].This paper explores a human-friendly interactivesystem, based on the harmonious symbioticcoexistence of humans and robots. A robotic cane isan interesting and excellent media that can be used toexplore various aspects of human-robot interactions. Itopens up the possibilities for truly human-friendlymachines. Firstly, the cane is an extension of thehuman body, allowing a point where the users’intentions and close interactions can be investigated.Secondly, if intelligence is fundamentally based oninteraction and doing the right thing in the real world,then sensing is the most essential component becauseit is what allows robots to perform in an intelligentmanner. Human sensing is dominated by sight. It isinteresting to see what happens if sight is impairedand how robots can work things out together withhuman partners in this situation [1,4,5].Our goal is to develop a robotic aid system capableof interacting with its environment and eventuallywith humans. Various methods for implementingappropriate cooperative recognition, planning, andcourse of action have been investigated [1,5,6].We outline a set of hardware solutions and workingmethodologies that can be used for successfullyimplementing and extending the interactivetechnology to coordinate humans and robots incomplicated and unstructured environments. Theissues discussed include methodologies for humanrobotinteractions, design issues of an interactiverobotic cane, hardware requirements for efficienthuman-robot interactions, and behavior arbitrationmethodologies for navigation planning.
2. ROBOTIC CANE “ROJI”
The successful and widely used travel aids for theblind are the white cane and the guide dog. Electronictravel aids are also used, but not so extensively. Bytaking advantage of the white cane and the guide dog,we have been involved in developing a robotic aidsystem named “RoJi” as shown in Figs. 1 and 2[1,7,8].The proposed robotic cane, “RoJi,” consists of along handle, two steerable wheels, a sensor unit that isattached at the distal end of the handle, and a userinterface panel. Much like the widely used white cane,the user holds these robotic canes in front of him/herwhile walking. The handle is painted in white tomimic the conventional white cane. The sensor head,which is mounted on a steerable, powered, twowheeledsteering axle, includes three infrared sensors,two antenna type contact sensors, and a sonic scanner.The robotic cane named “RoJi,” makes independentdecisions concerning the path it takes. However, theuser and the cane may wish to go in differentdirections. The normal operating mode of theconstructed robotic cane must include an overridefeature to allow the visually impaired individual to bein control when the need arises. We have tried to builda system that is capable of interacting with itsenvironment and eventually with humans.The contents of the robotic cane system and howthese components are used to provide desiredfunctional capabilities are described.2.1. Control and action module“RoJi” utilizes an on-board PICBASIC 2000microcontroller to process the sensor information andthe timer counts, and also to generate control pulsesfor the DC servo motor. The PICBASIC 2000 can beeasily programmed in Basic with the built-in analogand digital interface functions. The μ-controller isconnected to thirty-two digital inputs/outputs, twoPWM (pulse width modulation) outputs, one highspeedpulse counter, eight 10-bit A/D (analog-todigital)inputs and a 64 Kbyte flash memory [9].“RoJi” is driven and steered by two poweredmotors. Therefore, it can guide the blind individualautonomously with sufficient power. The two frontsteeringwheels of the cane are controlled independentlyFig. 3. Sensors for “RoJi.”by separate DC motors. One unpowered wheel in theback stabilizes the cane’s stable structure, enablingsharper turns.2.2. Sensing module“RoJi” has a sensor head unit that is attached at thedistal end of the handle to detect obstacles. Thissensor head unit consists of an active ultrasonic sensorunit, three infrared sensors, and two antennas forcontact sensing. A sensor unit, as shown in Fig. 3, ismounted above the guide wheels of “RoJi.”The active sensing unit mounted above the guidewheels has an ultrasonic sensor driven by a RC motor,as shown in Fig. 3. The unit scans the area ahead of itto efficiently detect obstacles or safe paths and canreduce the missing areas caused by narrow coveragedue to the fixed sensor arrangement. Its scanningangle is ±90° wide. The ultrasonic sensor detects anyobstacle up to a distance of 250 cm and within a rangeof ±30°.An array of three infrared sensors is utilized todetect the convex obstacles blocking the cane’spathway. These three sensors are arranged in a semicircularfashion, 30° apart from each other, and candetect any obstacle within a distance of 60 cm. Thisarrangement is based on the user’s shoulder width.Including the size of the robot platform with theradius of 18 cm, we can assure a safe pathway up toapproximately 80 cm wide.For short-range coverage, additional antennas forcontact sensing could effectively complement theseinfrared sensors and the active sensing unit. Limitswitches, angle detectors, or torque sensors attachedto these antennas quickly and easily detect dynamicchanges in the environment when the potentiometerscontact a surface or a bend [10]. Limit switches wereutilized in the earlier prototype of robotic canes [11].In place of these, potentiometers are utilized for“RoJi” [1]. Each antenna is arranged between theinfrared sensors. It must be made of flexible materialsso as not to interfere with navigation when it touchesthe surface to detect obstacles. Steel was selected for“RoJi,” and the length of each antenna is 22 cm.These antennas are connected to the potentiometer to detect sudden irregularities of the surface and to reactappropriately to them. Antennas are especiallysuitable to take advantage of the visually impairedindividual’s superb tactile information processingcapabilities.2.3. Operator interface moduleA miniature panel that can be operated with thethumb allows a user to specify a desired direction ofmotion, as shown in Fig. 4(a). This operator interfacemodule also allows the user and the cane to shareinformation to cooperate with and compensate foreach other. The user receives the cane’s obstacleinformation as different tones of audio signalsproportional to the distance. It then triggers the propernavigational command buttons.The user can point the active sensing unit in his orher viewing direction. The active sensing unit of“RoJi,” as shown in Fig. 4(b) can track the user’s headmovements by utilizing a gyroscope sensor attachedto his/her head. A semiconductor-type gyroscope,muRata’s ENV-05S, is utilized [12].Users can recognize the distance and the directionof the obstacles based on audio and gyroscopeinformation. Once the navigational path is determined,the user can control the robotic cane by pressing thebuttons on the interface panel. The operator interfacemodule contains four command buttons: “GoStraight,” “Turn Left,” “Turn Right,” and “Stop.” Asmall motor attached to the panel vibrates back to theuser, depending on the ground information. Also,speakers carry alarm sounds to alert the user. The canebenefits from the user’s flexible decision capabilities(a) A miniature user-interface panel.(b) Gyroscope sensor for head tracking.Fig. 4. User interface for “RoJi.”and his/her tactile and auditory information processingcapabilities.
3. NAVIGATIONAL PLANNING
Shared navigational planningPedestrians usually detect obstacles and navigateproperly according to visual information. The user’sdecisions for proper navigation are based on the user’sa-priori, intuitive and heuristic mental processing,instead of precise computations. In contrast, mobilerobots’ lack of intuitive nature requires precisegeometric information relating to the obstacles andgoals. A blind or visually impaired traveler assisted bya white cane and/or a guide dog only requireapproximate geometric information concerning theobstacles and goals.Robotic aids, such as robotic canes, needcooperation between humans and robots. Theconstructed robotic cane is self-steering, based uponthe interpretation of its input. The steering commandof the robotic cane is based on the absence or presenceof obstacles sensed by photo sensors. Without ashared control framework in place, the cane wouldwant to avoid objects such as stairs, doors, or chairsthat the user might wish to use. The operator’sdecision making must be included. Clearly the normaloperating mode of the robotic cane must beoverridden to allow the person to be in control whenthe need arises.The proper control commands for such a systemwould be a combination of autonomous commandsand commands from humans, rather than eitherautonomous commands or solely human commands.Our robotic cane has two control modes, the robotcontrol mode (RCM) and the user control mode(UCM). This shared control framework

download full report
http://googleurl?sa=t&source=web&cd=1&ve...98-405.pdf&ei=BeJ-Tby2M8qPcf6ZufMG&usg=AFQjCNH_lLIxNM-KdRWTiS7PvfeOG5PEBw