Intelligent Coaching Space

Project Acronym: 
01.10.2013 until 31.12.2018

ICSPACE explores how to provide intelligent coaching in sports training, motor skill learning, or physical rehabilitation. The goal is to develop a VR-based "Intelligent Coaching Space" that enables novel ways of adaptive, online feedback in a closed-loop interaction and training system. Leveraging fundamental cognitive principles, perceptual processing and movement understanding, ICSPACE combines state-of-the-art motion tracking and analysis, neuro-cognitive diagnosis, and multi-sensory (visual, auditory, haptic) feedback in various forms, including virtual mirrors, online verbal feedback, or instructions and demonstrations by virtual coaches.

This novel approach combines two key ingredients of successful coaching:
(1) A deep grounding of intelligent coaching in fundamental cognitive principles, perceptual processing and movement understanding;
(2) Adaptive support in an online, closed-loop sensorimotor learning and interaction system.

Scientific Goals: 

We aim to investigate the sensorimotor, cognitive, and social mechanisms underlying successful coaching, and to explore how cognitive interaction technology can provide coaching assistance that achieves, and perhaps even exceeds, the efficacy and acceptance of human coaching. Effective coaching for sensorimotor learning and skill acquisition requires having a number of challenging desired characteristics:

Competence: Advising an appropriately structured training plan with suitable exercises.
Attentiveness: Recognizing and responding to the trainee's performance and progress over the course of one or multiple training(s).
Instructiveness: Correcting errors made by the trainee by means of adaptive and informative feedback.
Comprehensibility: Giving understandable instructions and demonstrations of the movements to be practiced.
Engagement: Communicating empathically with the trainee in order to ensure understanding and motivation.
Personalization: Meeting the specific requirements of the trainee regarding level of expertise, physical limitations, exhaustion, affective or motivational state, etcetera.

To advance the use of Cognitive Interaction Technology in effective coaching and deep assistance applications, we intend to create an Intelligent Coaching Space that supports human trainees in learning and practicing sensorimotor skills through adaptive and personalized instruction and tutoring in an online, closed-loop sensorimotor learning and interaction system.

Work Packages: 

The work plan of ICSPACE consists of six work packages, each of which will address a main component of the targeted intelligent coaching environment. The work packages will run longitudinally through the project period. Their internal structures and progress are organized along the four milestone demonstrators, for which their results will be integrated and interwoven. The work plan comprises the following work packages:

Motion: Addresses questions concerning the recording, representation, analysis and interpretation of human motions.
Character: Creates different life-like virtual characters (e.g. from 3-D body scans) for the trainee avatar or the coach.
Feedback: Focuses on aspects of sensorimotor learning and multi-sensory feedback.
Dialogue: Concerns the verbal and nonverbal communication skills of the coach.
Integration and Evaluation: Provides an immersive virtual experiment environment and develops the software architecture.


The following milestones will be reached:
(I) Active coaching through multi-sensory interaction in a virtual world;
(II) interactive coaching through augmented online feedback in a virtual mirror;
(III) adaptive coaching by a real-time conversational virtual coach;
(IV) effective coaching by a personalized, intelligent coach.

Virtual World: During the initial installation of ICSPACE, we focus on tracking human motion, devising multimodal feedback technologies and evaluating sensorimotor feedback strategies to lay the foundations for the project. In this setup, the trainee’s motion is displayed through a coarse template avatar, which replicates the motions of the trainee in real-time.
Virtual Mirror: In the second demonstrator, we introduce two virtual characters: a virtual clone of the trainee based on an animated 3-D scan used to mirror the trainee’s performance, and an autonomous virtual coach that initially demonstrates exercises and gives feedback after the exercise has been performed by the trainee.
Virtual Coach: In the third demonstrator, the focus shifts from direct multi-sensory feedback to communicative feedback by the virtual coach. We will also evaluate whether modifications to the coach’s visual appearance, movement style, and demonstrations based on the appearance and capabilities of the human trainee have an effect on the trainee’s performance.
Personalized Intelligent Coach: In the final demonstrator, the coaching capabilities will include individualized planning of a long-term coaching program. At this stage, all levels of representation, from the biomechanical model up to mental structures by means of neurocognitive diagnostics, are included in the coaching process. The intelligent coach will employ coaching strategies adapted to the capabilities and the training progress of the trainee, and establish long-term motivation by social feedback.

Ausgewählte Publikationen des Projektes: