1999 Conference Proceedings

Go to previous article 
Go to next article 
Return to 1999 Conference Table of Contents


Albert M Cook
Faculty of Rehabilitation Medicine1 & Faculty of Engineering
Edmonton, Alberta, Canada

Kathy Howery
University of Alberta and Glenrose Rehabilitation Hospital
Edmonton, Alberta, Canada


The goal of this research is to investigate whether young children who have physical disabilities will be able to use a robotic arm for exploration and discovery. These children have difficulty in manipulating toys and other objects, and the robotic arm can provide an alternative method to accomplish these manipulation tasks. Using a robotic arm, children who have physical disabilities are able to engage in a manipulative activity involving a sequence of steps. They are able to experience, independently for the first time, the manipulation of real objects.


From infancy onward learn about their environment by exploration and discovery. Piaget termed the first two years of life the sensorimotor period (Thomas, 1992). During this stage of development, infants interact with their environment, learning from physical sensations and sensory input (Linder, 1990). Children learn the relationship between their own bodies and the environment through repetitive movement and exploring the environment through their senses. At the end of this sensorimotor period,children have a sense of 'object permanence', objects' positions in space and time, and causal relations among them.

They learn these concepts through active exploration and manipulation of their environment. Children with severe physical disabilities have limited opportunities to explore their environment as infants. Their lack of spontaneous, independent exploration and discovery of their surroundings may influence both their cognitive and social development. This lack of experience carries over into later childhood.

These children are often dependent on their parents or caregivers to assist them with interactions with their environment. Indeed, Brinker and Lewis (1982) suggest that the behaviour of children with physical disabilities may itself shape parents' choices of environmental exploration to only a few parent-defined situations. Thus even the choices for play provided by their caregivers may be very limited. As a result, an already restricted repertoire of exploratory abilities is further restricted by the parent-child dyad.

Children who are physically dependent cannot control their environment and have few opportunities for spontaneous exploration. They may begin to lose interest in the environment and develop 'learned helplessness'. Learned helplessness is the perception of persons that they cannot control the outcomes of their actions, and is characterized by passivity and avoidance (Todia, Irvin, Singer & Yovanoff, 1993).

Children with severe motor dysfunction may remain passive, dependent and socially inactive unless their specific developmental needs are recognized, and they are given active opportunities to explore their surroundings by alternative methods (Scherzer & Tscharnuter, 1990). Only then are they afforded the opportunity to experience mastery motivation, the opposite of learned helplessness, which occurs when children have an opportunity to experiment with a behaviour and receive positive reinforcement (Todia, Irvin, Singer & Yovanoff, 1993).

Assistive technology has given children with physical disabilities a way of exerting control over their environment. For example, Swinth, Anson, & Deitz (1993) showed that children as young as 6 months old were able to access a computer-controlled cause and effect program by hitting a single switch.

The use of robotics provides children with physical disabilities the unique opportunity to choose how they want to interact with their environment, and to exert some control over the environment, rather than having an activity forced upon them. Robotic systems also allow manipulation of real objects, (3 dim.) rather than computer graphics (2 dim.) for simulated manipulation.

Cook, Liu and Hosseit (1990) carried out a study to determine if very young children would interact with a robotic arm. Six disabled and three normal children, all less than 38 months in age, were used in the study. The system consisted of a computer for control and data collection, and a small robotic arm (Cook et al., 1988). Children used the robot arm as a tool by pressing the switch only when it was necessary to bring an object closer or to uncover a hidden object.

Fifty percent of the disabled children (all of those with a cognitive developmental age of 7 to 9 months or greater) and 100 % of the normal children interacted with the arm and used it as a tool to obtain objects out of reach. Gross and fine motor skill levels were less related to success in using the robotic arm than were the levels in cognitive and language areas.

Cook et al. (1997) extended the earlier work to focus on child-directed exploration and discovery. Their objectives were:

(1)To evaluate how children with severe physical disabilities use a robotic arm for exploration, and
(2) To determine the relationship between switch-mediated choices and the complexity of the resulting play tasks.

This feasibility study evaluated how children (aged 3-6) who have physical disabilities use the robotic arm for exploration. The robotic arm was used in an exploratory format in which the child was given access to various movements by pressing one or more switches.

A series of progressively more complex tasks reflecting both the increasing developmental level of the task and an increase in the number of switches used to control the arm was programmed for playback by the child. The child was shown how the robotic arm moved by a single switch activation to replay a stored movement and encouraged to press the switch. To maximize the experiences of the child, dry macaroni in a tub was used to provide both sensory and motor interactions for the child. The tasks used were:

(1) The arm dumped macaroni from a glass using one switch hit,
(2) The child controlled the arm to fill the glass and pick it up, and to dump it, using two switches
(3) The child caused the arm to move laterally and fill the glass with macaroni, raise the glass, and dump the macaroni into the tub using three switches.

Each session with a child was video taped for review. Observations included the child's actions and behaviours during the performance of the tasks. For example: looking toward the arm, looking at the switch, positive or negative (e.g. smiling or crying) response in anticipation of or in response to a robot action, restless- ness, attention to task. Results indicated great interest by the children in the robot and no apprehension toward it.

Children generally attended to task for significantly longer periods with the robot than with other activities (e.g., computer graphics programs). The children also appeared to understand the stored tasks by repeatedly executing them. Positive responses, looking toward the arm or switch and general high level of attention were closely related in time to robot switch activation by the child. The study reported here extended the earlier work to focus on understanding by the child of sequencing and the use of the robot arm for exploration.


The robotic arm used for these studies is the CRS A465 which can rotate about its base, flex and extend at the elbow and shoulder, extend, flex, supinate and pronate the wrist, and open and close the gripper. Using a teaching pendant, the arm can be moved through a desired movement or portion of a movement and the movement can be stored for later playback. The robot control computer, and electronic controller also record data on interaction with the robotic arm. In this study, we set out to determine whether the children (aged 4-7) who have physical disabilities would be able to understand a sequence of motor actions and to use them to find buried objects of interest. Three distinct tasks were programmed to generate specific robotic arm movements. Each movement was activated and executed by a single switch press. Dry macaroni in a tub was used to provide both sensory and motor interactions for the child. The tasks used were:

Task 1: The arm dumped macaroni from a glass using one switch hit.
Task 2: The child controlled the arm to
(1) dig an object out of the macaroni, and
(2) to dump the macaroni and object, using two switches.

Task 3: The child caused the arm to
(1) move laterally too a location where an object had been buried,
(2) dig the object out of the macaroni, and
(3) dump the macaroni into the tub, using three switches. The object was a plastic egg containing another object of interest to the child (e.g., finger puppet, small rubber stamp)

Our study focussed on determining whether the child understood the function of each switch and whether he or she could combine the switch actions into a three step process (Task 3). Four children, aged 6-7 participated in the study. All of the children had severe cerebral palsy which limited their ability to control their limbs.

None of the children were able to speak. Each child used left and right head movement for the first two movements (those in Task 2). The third switch (for Task 3) was activated by the hand for two of the children, behind the left elbow for one and by the right foot for one. Each session with a child was video taped for review. Observations included the child's actions and behaviours during the performance of the tasks and the type and number of prompts which were required in order for the child to successfully complete the task.


There was great interest by the children in the robot and no apprehension toward it. Children generally attended to task for significantly longer periods with the robot than with other activities (e.g., computer graphics programs). All of the children understood that hitting switch #1 dumped the cup and its contents. After one or two trials in which verbal and physical prompts were used, the children were able to activate the switch as soon as the cup was filled by the experimenter. The children reacted very positively to this task and looked toward the arm after hitting the switch in anticipation of the cup being dumped. This phase lasted an average of 8 trials, and it served as a "warm-up" both physically and to the task for the child.

Adding a second switch with a different function led to some confusion for the child. All of the children had to be physically prompted (moving the child's head until it contacted switch #2 (dig)) before they began to understand the task. After one or two physical prompts, each child learned to "dig" (switch 2) and then "dump" (switch 3) with only verbal prompts. Due to their severe physical disabilities, some of the children had uncontrolled body movements that led to some random switch activations (e.g., hitting the "dump" switch before the "dig" switch). Children took between 6 and 12 trials to demonstrate that they understood the two switch sequence tasks by completing it for three trials in a row with no prompting. When the third switch ("move") was added (Task 3), the children required differing levels of prompting to understand its function.

For one child, during one sessions, it was necessary to remove switches 1 and 2 and teach the function of switch 3 alone, then replace the other two switches. For the other sessions and for all sessions with the other children, the third switch was added to the first two. The child required both verbal and physical prompts in order to carry out the third part of the sequence ("move"). Children took more trials to understand this task, and each trial required more prompting. There were also many incorrect switch hits (wrong order). All of the four children demonstrated understanding of the three switch task, but the results were less conclusive. Many more prompts were required for this task, than for the other two tasks.


All of the children understood the first two stages (dump-1 switch, and dig/dump -2 switch) very well. After one or two prompted trials, the children hit the proper switch to "dig" prior to hitting the "dump" switch. For some of the children, the switches were sometimes hit out of order. This may be due to extraneous motor activity and/or lack of understanding of the task. Further research is required in order to distinguish between these two options. The three step (3 switch) move, dig, dump sequence also is appears to have been understood. There were many incorrect stich activations. wrong unfortunately Once again, this may be due to the extraneous movements or lack of understanding. In the current study, the computer was programmed to accept only the correct switch input.

Thus, if a child hit the wrong switch the arm did not respond. It is now important to decide if the extra switch hits are accidental due to uncontrolled movements or due to lack of understanding of the task. In order to make this determination we will use an alternative control mode in which all switches are active at all times, and the child will cause the arm to respond no matter which switch is hit.

For example, if the child hits switch 2 first the arm will attempt to dig, even though it is not positioned near the buried object. This will allow us to explore what the child's reactions are to this anomaly and whether the child attempts to correct his or her actions to obtain the correct sequence.

While all children could correctly sequence the actions to complete the entire task in multiple action tasks, the number of sessions and trials to reach this level varied across children. Children were much more motivated to learn how to use the robot and they kept their attention focussed for longer periods of time in contrast to simple toys which do not generally allow the child to move beyond cause and effect relationships and computer programs which are not as concrete in relation to object manipulation and sequencing of tasks. Children were able to put two operations together to complete a task. This gave a unique "window" on their ability to sequence since most were very physically limited and did not speak. The robot arm also gave the children the opportunity to interact with the investigators by "handing" objects to them and choosing which objects to be buried. This is a preliminary study leading to the provision of more general manipulative learning opportunities for the children. We also plan to involve two children in a play task together using the robot arm.


Brinker RP & Lewis M (1982). Discovering the competent handicapped infant: a process approach to assessment and intervention. Topics in Early Childhood Spec. Educ., 2(2), 1-15.

Cook AM, Hoseit P, Liu KM, Lee RY & Zenteno CM: Using a Robotic Arm System to Facilitate Learning in Very Young Disabled Children, IEEE Trans Bio. Med. Engr., BME-35: 132-137, 1988.

Cook A, Howery K, Darrah J, Gu, Kostov A, Adkins R, Heaton E &Meng M, Robot-enhanced Interaction for Children with Disabilities, Proc. 1997 ARATA Conf.,1997.

Cook AM, Liu KM and Hoseit P: Robotic arm use by very young motorically disabled children, Assist. Technol., 2:51-57, 1990.

Linder TW (1990). Transdisciplinary Play-based Assessment; Functional Approach to Working with Young Children. Baltimore: PH Brooks.

Scherzer AL & Tscharnuter I (1990). Early Diagnosis and Therapy in Cerebral Palsy. (2nd ed). New York: Marcel Dekker.

Swinth Y, Anson D, Deitz J (1993). Single-switch computer access for infants and toddlers. American Journal of Occupational Therapy, 47 (11), 1031-1038.

Thomas, RM (1992). Comparing Theories of Child Development (3rd ed). Belmont CA: Wadsworth Publishing.

Todia B, Irvin LK, Singer GHS, & Yovanoff P. (1993). The self-esteem parent program. IN Singer GHS & Powers LE (eds) Families, Disability, and Empowerment. Toronto: Paul H Brookes.

Go to previous article 
Go to next article 
Return to 1999 Conference Table of Contents 
Return to Table of Proceedings

Reprinted with author(s) permission. Author(s) retain copyright.