Manus Quantitative Evaluation

Proceedings of the ICORR 2003 (The Eighth International Conference on Rehabilitation Robotics), 23-25 April 2003

Manus Quantitative Evaluation B. Abdulrazak1,2, M. Mokhtari 1,2, B. Grandjean2 1 Institut National des Télécomunications (INT), Evry, France. 2 INSERM-U483, University Pierre & Marie Curie, Paris, France. E-mail: [email protected] Abstract This paper describes our research activity on the integration of a robotic arm in the environment of persons with disabilities. People who have lost the abilities to use their proper arms to perform daily living task could use an adapted robot to compensate, even partly, the problems of objects manipulation generated by their handicap. Many robotized systems have been developed, however the manipulation of such robots by the disabled people is not always obvious. To meet the users needs it is necessary to involve the users in all steps of product development: design solution, prototyping the system, selection of user interface, and testing in real conditions. In this paper, we provide original results on the evaluation of the new Manus software architecture called Commanus in reference to the European Commanus project terminated last year. A quantitative evaluation method has been developed and used to provide accurate data on the usability of Manus robot with the integration of new control modes. This work correspond to 9 months of evaluation recording with about 19 quadriplegics patients, mainly having spinal cord injuries and muscular dystrophies, from the rehabilitation hospital of Garches. Introduction The main objective of our research work is to determine the different factors that could influence the human-machine interaction in term of assistive robotics for people having severe motor disabilities. Our contribution consists on providing quantitative evaluation methods which provides accurate data on the usability of a system, such as the Manus robot. Robotics for disabled people appeared in 1985 in France with the Spartacus project [3]. This has permitted to highlight several points regarding the Human-Machine interaction in case of handicapped people using robots. This research work served as a reference in the robotics development which followed [6], in particular the AfmaMaster workstation and Manus robot. The French Muscular Dystrophy association (AFM), in close collaboration with our team, has launched an operation intended to provide a Manus robots to some handicaped people living in their homes. The aim of this approach was to verify if a Manus robot could give substantial help to the

users in their daily environment. The implication of our team in this operation consists on integrating our quantitative evaluation method to get accurate information about the usability of Manus robots in real conditions at home. This experimentation has permitted to point out many points regarding the adaptability of this robot to handicapped people [4]. In other words our contribution for designing adaptable, configurable and personalized robot system based on the Manus robot (Commanus Project) allowed building a new Manus control system adaptable in function of user [2]. In this paper we present preliminary evaluation results obtained with the latest version of Commanus control system. Users needs analysis As it was noticed in former evaluation performed by our team [4], users spend more than 50% of the task duration seeking for the good strategies to reach the target or seeking for the appropriate control action. The use of Manus arm in daily living with the commercialized command architecture implies always that the user has to perform repetitive actions to realize the same tasks, for example; when the user targets an object, he performs manually using a succession of actions on “Cartesian” or “Joint” modes to move the robot toward the target, when this first phase is finished, the user performs manually also, the necessary commands to grasp the desired object (second phase). When the object is small or the task require precise, for example when user wants to insert a floppy disk in a computer, the second phase could be complicated and needs many actions to adjust the position of the floppy. Our approach is to propose assistance to the user when using the Manus in an open environment. We developed a new control software architecture which allow the handicapped user to perform the same sequence of commands as a global automatic movement of the robot stored in a gesture library. The new control modes proposed provides semi-autonomous control of Manus that permits to decrease the number of actions necessary to perform tasks. The new system permits, in one hand, to reduce manipulation problems that users meet during complex tasks, and on the other hand, to solve the problems linked to the user-interface. With these new functionalities, we noticed a reduction of task duration and the decrease the number of actions necessary for complex task.

Evaluation method

Hardware architecture

The aim was to develop original methods based on quantitative evaluation to analyze accurate data on the usability of the Manus robot and particularly the contribution of the new added modes. The idea was to record all the actions performed by the users on the input devices. The log file generated contains several parameters, such as all commands performed by the user on the input device, the processing time of the robot, the corresponding mode, and the robot gripper orientation and position coordinates. This method allow us to determine several key parameters on the usability of Manus, such as, the time spent using the Manus according to each control mode, how many actions processed in each mode, and how many warnings and error messages has been generated.

The new control box is based on a PC-104 computer card; this system was planned for the procedure of evaluation. Due to hardware integration delay in this portable computer box, we decided to proceed with the evaluation using a standard desktop computer which integrates the Commanus control software (Fig 1). This system supports as input control devices: a 16-digit keypad, a 3D joystick, a trackball and any type of mouse. A screen (visual feedback) displays a user-friendly Graphic Interface for mouse control or with scanning devices controlled by one switch. This screen represents a devices input that helps the user to send commands to Manus. Users can also find, Manus status and warnings messages that are helpful during manipulation. Visual feedback was used during the whole evaluation period. Configuration software tool “OT” allows the occupational therapist to configure easily input devices with different menus containing activities associated to action commands of Manus. During the evaluation, different parameters where on line. This method allows to know how much time the user spends in each mode, how many actions he makes in each mode, and how many warnings and error messages occured during manipulation. Data analysis allowed us to compare the use of the Manus among different type of tasks and among different endusers.

Hardware & Software organization Manus is a tele-manipulator robot mounted on an electric wheelchair. Its objective is to favor the independence of severally handicapped people who have lost their upper and lower limbs mobility, by increasing the potential activity and by compensating the prehension motor incapabilities. Manus is a robot with six degrees of freedom, with a gripper in the extremity of the arm which permits capturing objects (payload of 1, 5 kg) in all directions. It is controlled by a 4x4 buttons keypad, a joystick, a mouse with the latest version available in our lab, and also with a touch screen in the near future. Theses input devices give the user the possibility of handling Manus, and the display unit gives the current status of the robot [1]. The main advantage of the Manus is that it can perform tasks in a non-structured environment which correspond, in general, to the real environment of the end-users. However the manipulation of such assistive device by the disabled people is not always obvious [4].

Fig 1: System to collect data from quantitative evaluation

Software command architecture The Manus software architecture allows us to choose many modes in order to offer several possibilities of controlling the arm [7]. The basic software architecture, called Manus modes, has three different control modes: the Main Mode, the Cartesian Mode and the Joint Mode. Regarding the evaluation results, we have developed new command architecture, called Commanus modes, and implemented several extra modes beside the Manus modes to meet the user’s needs. Four new control modes have been developed and integrated to the software architecture of Manus robot: The first additional mode is the Record & Replay Mode which allow the user to record a specific gesture for replay in case of repetitive task. The second one is the “Pilot mode” which allows handling Manus robot in the direction of the gripper following the main axis. This mode has been developed mainly when using a 2D joystick to control the robot when practicing robot forward action in the direction of the target, which is similar to the human approaching movement to grip an object. The third mode is the Relative mode which allow movements of the gripper, usually where it is near the target, to perform small defined steps relatively to the target position and to the current position and orientation of the gripper. This mode is useful during task requiring high accuracy, such as inserting a video tape in a VCR [5].

Preliminary results These preliminray results correspond to 9 months of evaluation and about 37 hours of recording with quadriplegic patient, mainly having spinal cord injuries and muscular dystrophies, from the hospital of Garches. 19 Persons took part in our evaluation, 2 muscular dystrophies patients, 13 hospitalized spinal cord injuries patients, 2 patients with cerebral palsy and 2 patients with Locked In Syndrom. One of the spinal cord injuries patients has more than 2 years of its accident, he lives in a specialized institution. This patient had taken part in the evaluation of the first version of Manus, since this time, he expressed his wishes to use the arm in situation of real life. 1 person having muscular dystrophies is living in his home with his family and had taken part in a former evaluation of Manus, he had acquired a Manus robot embarked on a wheelchair for one year. The participation of these two persons in this evaluation with Commanus allowed us to compare with the commercialized version of Manus. The average age of the users taking part in this series of e valuation is 32 years, 2 people more than 50 years and a you ng spinal cord injuries patient of 10 years old. Modes and time of use The first graph (Fig 2) shows the time repartition during the whole evaluation duration of 37 hours (134.360.392ms): Mode (Execution Tim e) Joint relative 234 227,00

2 479 878,00

Pilot

3 124 476,00

Cartesian position

3 497 029,00

Joint velocity

5 142 332,00

Cartesian velocity NOP

Mode (Nb actions) Pilot (15,0%) Cartesian relative (4,1%) Cartesian position (3,4%)

Joint velocity (23,7%) Cartesian velocity (51,9%) Joint position (0,8%) Joint relative (1,1%)

Fig 3: Number of action repartition Users’ manipulation on the input device generated 633 events without any robot activity (keypad buttons or joystick event without function). The Robot generated 92 warnings messages (robot in deadlocking configuration, limit of working space reached…). To recover deadlocking configuration, the user have recourse to the joint mode to ovoid restarting the system. During the evaluation all users had to perform the same tasks, and to follow the same scenarios. Different parts have been performed according to different control modes: P-01: part using the basic modes and follow with; P-02: P01+Pilot mode, P-03: P-01+Point to point mode, P-04: P01+Relative mode and P-09: free scenario (all modes). We remarked that within the parts using the new modes, users needed less number of actions and less time to perform the tasks, which mean that Pilot mode (Fig 4 & Fig 8), relative mode (Fig 6 & Fig 9) and Point-to-Point mode(Fig 5 & Fig 10), contributed in the reduction of the number of actions and to the reduction of duration.

Cartesian relative 868 100,00 Joint position

Actions number The representation in term of events or actions performed on the input device is shown below (Fig 3). The whole recording time correspond to 9715 Events actions sent to the robot.

10 837 469,00 108 176 881,00

Mode contribution by number of actions Fig 2: Time repartition for control modes The “NOP” time (no actions or rest time) is considerable and represents 80,5% of the total duration. But we have to make the distinction between three types of rest times: - a no action time (when the user takes a real rest without switching off the Manus), - a cognitive time (when the user is thinking on the sequence of action he plans), - a physiological motor time(the physiological time necessary to execute a movement with hand or finger). The “Cartesian velocity mode”, corresponding to the Cartesian Mode, is the most frequently used (8,1% in time) and could be processed with three different speeds: slow, medium and high speed. The Cartesian velocity mode with medium speed (2) is the most frequently speed used (53% in time) where the “Cartesian mode” in high speed (3) is not really used (1,2%), it is only when users want to do large movement of the arm robot.

P-01 / P-03

P-01 / P-02 U-01

U-01

U-02

U-02 P-01 P-02

P-01 P-03

U-09

U-09

U-14

U-14

U-15

U-15

Fig 4: Pilot

Fig 5: Point To Point P-01 / P-09 -> Task T-04

P-01 / P-04 P-01 P-04

U-02 U-14

Fig 6: Relative mode

U-01 P-01 P-09

U-02 U-14 U-15

Fig 7: Commanus modes Using Manus within all new modes at the same time; also, need less number of actions and less of time to perform the tasks (Fig 7& Fig 11), which confirms that the new modes contribute in the reduction of the number of actions and to the reduction of execution time.

Mode contribution by execution time Past Tim e (P-01/P-04)

Past Tim e (P-01/P-02) U-01 P-01 U-02 P-02 U-09 U-14 U-15

U-01 P-01 U-02 P-04 U-09 U-14 U-15

Fig 8: Pilot mode

Fig 9: Relative mode

Past Tim e (P-01/P-03)

Past Tim e (P-01/P-09) -> Task T-04

U-01 P-01 U-02 P-03 U-09 U-14 U-15

P-01 P-09

U-01 U-02 U-09 U-14 U-15

Fig 10: Point To Point mode Fig 11: Commanus modes

developed, this tool allows the user to choose device and change the configuration mapping of actions inside the modes as he seems appropriate. Evaluation of the contribution of the personalization of devices has been started; the results can be presented in the future. The continuation of this research work is insured through a new European project AMOR which should start in March 2002 and which is a logical continuation of Commanus Project. The development realized during Commanus will lead to a new command architecture for Manus which will be integrated in AMOR project. The aim is to propose a new generation of Manus robot taking into account the users requirements. Acknowledgments

Discussion During evaluations, we have noticed that: ♦ The standard relation between movement and mode shown that Cartesian, Joint or Pilot modes where generally used to carry out movements of high amplitude. The Relative mode remains adapted for the final phase, for example to insert floppy in a PC. ♦ The visual feedback contributed to simplify the training phase, this was concretized by the reduction in the training duration, a session of 20 minutes was sufficient for the majority of the users in contrary to several sessions previously. ♦ Use of the OT configuration tool, dedicated to define the mapping control on the input device, by the occupational therapists remained appreciable by the users. We noticed that it was necessary to carry out a series of input device’s configurations, on average from 2 to 5 iterations to lead to a personalized version for each user. ♦ The training of the Pilot Mode remains difficult to realize because it forces to imagine a virtual reference fixed on the gripper. But for those who succeeded this step, expert users, it permitted to decrease the execution time of the tasks. Even, some users found it more intuitive in comparison with the movement of the hand. ♦ These results confirmed our former evaluations on the favorable contribution of the Point-to-Point mode. Conclusion The evaluation of the new architecture allowed us to bring some improvement to the system. The first trials with disabled patients showed their interest regarding the new added modes. The preliminary results obtained show the real contribution of new command architecture modes. More evaluations in real life conditions, with the help of disabled people, are necessary to test all the new functionalities offered by the proposed new system. This new architecture allowed plugging many devices input. Users have choice of the devices input. With this architecture a simple and user-friendly “OT” tool is

The authors would like to thank the users who have participated actively in this évaluations and also C. Dumas and I. Laffont from Garches hospital. Funds for this project were provided by the AFM (French Muscular dystrophies Association) and by the European Commission. References 1) B. Abdulrazak, M. Mokhtari, B. Grandjean. (2003): «Assistive robotics for severely disabled people: The Commanus project », AMSE, Journal of the Association for the Advancement of Modelling and Simulation Techniques in Enterprises, special edition HANDICAP. 2) B. Abdulrazak, M. Mokhtari, B. Grandjean, C. Dumas : « La robotique d'aide Aux personnes Handicapées, le projet Commanus», Handicap'2002, la deuxième conférence, “Pour l’essor des technologies d’assistance”, Porte de Versailles, Paris, France, pp. 89-94. 2002. 3) R. Chatila, P. Moutarlier, N. Vigouroux, « Robotics for the Impaired and Elderly Persons. IARP Workshop on Medical Robots». Vienne, Autriche. 1-2 Oct. 1996. 8p. 4) J. Heidmann, C. Dumazeau : « Evaluation du robot Manus par des personnes lourdement handicapées », Rapport interne Handicom 1999. 5) M. Mokhtari , B. Abdulrazak, R. Rodriguez, B. Grandjean. (2003): «Implementation of path planner to improve the usability of robot dedicated to severely disabled people», 2003 IEEE International Conference on Robotics and Automation, Taipei, Taiwan. 6) M. Mokhtari, N. Didi, A. Roby-Brami, « A multidisciplinary approach in evaluating and facilitating the use of the Manus robot. IEEE International Conference on Robotics and Automation », ICRA’99, Detroit, Michigan. May 1999. 7) J. Vertut, P. Coiffet., «Les Robots ». Tome 3a : Téléopération, Evaluation des technologies. Ed. Hermes. Paris, 1984.