Log in / Register
Home arrow Computer Science arrow Robotic Assistive Technologies: Principles and Practice

Machine Learning and Early Intervention

Learning from clinician and parental input (which specific goals need to be met) and incorporating these data into decision making is crucial for the success of a SAR. Using this knowledge in real-world environments to elicit a desired response from the child can significantly improve clinical outcomes. For example, a clinician working to elicit vocal engagement behaviors from a child can incorporate information about how the child responds to physical closeness to the SAR, and the parents’ knowledge of the types of output a child responds to can be used to initiate SAR-child engagement (Robins, Dautenhahn, and Dickerson 2009).

Adaptability through machine learning makes the SAR more capable of working one on one with the child, so it may be used in home therapy. Clinical intervention techniques can be used as input for the robot to make decisions. Using reinforcement learning allows the robot to focus on the overall goal for the child, no matter what it may be, and interact with the child in a way to help them reach that goal. Using machine learning, the robot can emulate clinician behaviors following objectives set by the clinician until they reach the desired goal. Because the clinician sets the pattern with the robot, these play activities can be used to reinforce clinical intervention sessions (Allievi et al. 2014; Basu and Clowry 2015).

The SARs that are able to sustain engagement can be a valuable therapeutic tool for children with complex disabilities. The clinical target determines which characteristics would be most beneficial for the SAR, as each child is different. For example, one should keep in mind any auditory, visual, spatial perception, or language impairments, as each of these has an effect on their interaction (Michaud, Duquette, and Nadeau 2003). Each child has his or her own individual needs, so the SAR should either be relatively customizable or cover as large a range of impairments as possible. The remainder of this section provides a comparative example by describing design considerations for children with ASD and children with CP. Children with ASD are considered because studies have shown that SARs can be effective therapeutic tools for these children. This is then compared to CP to show that SARs can have both major similarities and differences between various developmental disorders (Robins et al. 2012; Syrdal et al. 2014).

Using SARs to bridge the gap between no communication and normal communication has been a frequently explored therapy tool (Bosseler and Massaro 2003; Boucher 2003; Feil-Seifer and Mataric 2009a; Ferrari, Robins, and Dautenhahn 2009; Light et al. 1998; Robins, Dautenhahn, and Dickerson 2009; Tapus et al. 2012). For these children, maintaining their attention and prompting them to interact a certain way has been accomplished through significant design considerations, including the following:

  • 1. Attraction: Children with ASD often respond better if they can see electrical or mechanical components, so creating a robot that exposes some of the inner mechanisms can be beneficial (Salter, Davey, and Michaud 2014). Conversely, designing a robot with human features can also attract the child’s attention and help the child adjust to interacting with people while interacting with a robot (Robins, Dautenhahn, and Dickerson 2009).
  • 2. Adaptability: Each child will respond differently, so having a robot that can adapt to personality types would be the most beneficial method for interaction (Salter, Davey, and Michaud 2014).
  • 3. Maintaining Interaction: Keeping the child engaged for an extended period of time allows the robot to have greater therapeutic benefit. If the child engages the toy but quickly leaves it or becomes disinterested, the child will be less likely to benefit from the SAR (Robins et al. 2012; Salter, Davey, and Michaud 2014).
  • 4. Routine Movements: Children with ASD can sometimes become locked in to a particular repetitive movement. A SAR can counter this by mimicking this routine movement and slowly prompting them into more typical behaviors (Salter, Davey, and Michaud 2014).
  • 5. Simplistic Design: Robots are much more simplistic than humans when it comes to communication, which may be a contributing factor regarding successful robot interaction for children with ASD. Keeping the design relatively simple may help the child focus on communication or performing a given task to reach a specified goal (Salter, Davey, and Michaud 2014).

Like ASD, CP encompasses a wide range of developmental delays. However, in addition to focusing on mediating social and communication interactions, children with CP can use SARs to reach physical, sensory, and cognitive developmental milestones and receive therapeutic interaction that they may not be regularly able to access. For this reason, the design considerations for this type of SAR may include the following:

  • 1. Attraction: SAR therapy for children with CP can occur at a very early age, so the robot needs to reflect the cute, endearing models that attract young children. This may include bright colors, soft fur, large eyes, or blinking lights (Howard 2013).
  • 2. Adaptability: Like with ASD, a SAR for children with CP will need to be able to adapt to a child’s individual needs. This adaptability can help maintain interaction while providing the best therapeutic experience for the child (Calderita et al. 2013).
  • 3. Repetitive Movement: To reach a developmental milestone, the child will most likely need to have consistent practice making a designated movement. This can be accomplished by designing a SAR to encourage a child, through play, to repeat this movement until the child reaches a specified therapeutic goal, such as reaching across midline (Howard 2013; Robins et al. 2012).
  • 4. Simplistic Design: Making the robot endearing to children will be important, but the robot should not become overwhelming. The design should be simple for the child as well as for the parents so they may use the SAR as a therapeutic tool at home. Making a complicated device could end up frustrating the child and parents, preventing them from accessing the benefits of the robot (Ljunglof et al. 2011).

Comparing just a few of the many available design considerations for the two populations shows that, in both cases, SARs need to be individually customizable, even while targeting a specific audience. Not only do specifics of developmental disorders have a significant range, but also the differences between individual children of any disorder are just as vast.

Robotics Safety Standards

In addition to the design considerations mentioned, SARs must be safe to use. Current guidelines for robot safety include the International Organization for Standardization standard, ISO 13482 International Standard: Robots and Robotic Devices—Safety Requirements for Personal Care Robots (please refer to Chapter 1). The proper selection of an effective robotics safety system is based on hazard analysis of the operation involving a particular SAR. Among the factors for consideration in such an analysis are tasks a robot is programmed to perform, the startup and the programming procedures, environmental conditions and location of the robot, requirements for corrective tasks to sustain normal operations, human errors, and possible SAR malfunctions.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science