Objectives
Develop and evaluate the multiparty communication and collaboration capabilities of the KASPAR robot, as well as use the robot as a mediator for enhancing the collaboration skills of children with ASD. The design, development and evaluation of these capabilities will be grounded on child-child and child-robot interactive scenarios that constitute the third use case of BabyRobot. Long-term experiments (case study evaluations in schools) with the ASD user population will be performed and the data collected will be evaluated by ASD experts and educators to verify the progress in the collaborative abilities of users. Most importantly the collaborative capabilities of the robot will be enhanced via reinforcement learning and goal formation on the fly using the intention reading and shared intentionality modules.Scenarios will be developed, implemented and tested in light of the specific education and therapeutic needs of the children involved in the studies.
Description of Work
- Use Case 3 Specification and Data Collection Protocols: This task will provide the overall specification of use case 3, using the humanoid and minimally expressive robot KASPAR that has been designed by UH specifically for social interaction. Use case 3 will focus on multiparty interaction games. Data collection and analysis will be specified. The experimental protocol will be specified, in particular with regards to how improvements of children with autism’s collaborative skills can be demonstrated.
- Learning Collaboration Skills and Cognitive Development: The goal of Use Case 3 is to allow the robot to learn from and adapt during the interaction with the children. The models developed for joint attention, common grounding and especially for shared intentionality will be adapted to the collaboration scenario in order to both predict and assess the collaborative skills of a child. Cognitive models for each individual child participant attention, common grounding and intentionality will also be built for assessment purposes.
- Development of Multiparty Collaboration Scenario: A multiparty collaboration scenario will be implemented. This requires the design and building of a particular multiparty game setup, possible hardware modifications to the KASPAR robot as needed for the scenarios, and software development realizing the scenario. In addition, we will investigate adaptive learning (e.g., reinforcement learning) algorithms for goal formation and collaboration on the fly.
- Integration and Evaluation for Use Case 3: The developed algorithms will be evaluated in terms of improved performance of the collaborative and communicative nature of the interactions of the robot with the children. The top- performing algorithms will be selected and integrated into the KASPAR platform. The robotic platform will be evaluated using objective and subjective metrics. Individual modules of the platform will also be evaluated using objective metrics on annotated data.
- Evaluation: Working with ASD therapists and educators: We will work closely with ASD therapists and educators as part of their ongoing work into robot-assisted therapy for children with autism. They will be also involved in the evaluation of the increasingly advanced prototypes of the use cases. The data analysis (questionnaires and interviews with professionals, behavioural data of children with ASD) will be carried out on an ongoing basis. Behavioural data of children with autism will be analysed using the Observer software and experienced coders.