An Augmented Reality Serious Game for Learning Intelligent Wheelchair Control: Comparing Configuration and Tracking Methods.

This work proposes an augmented reality serious game (ARSG) for supporting individuals with motor disabilities while controlling robotic wheelchairs. A racing track was used as the game narrative; this included restriction areas, static and dynamic virtual objects, as well as obstacles and signs. To experience the game, a prior configuration of the environment, made through a smartphone or a computer, was required. Furthermore, a visualization tool was developed to exhibit user performance while using the ARSG. Two user studies were conducted with 10 and 20 participants, respectively, to compare (1) how different devices enable configuring the ARSG, and (2) different tracking capabilities, i.e., methods used to place virtual content on the real-world environment while the user interacts with the game and controls the wheelchair in the physical space: C1-motion tracking using cloud anchors; C2-offline motion tracking. Results suggest that configuring the environment with the computer is more efficient and accurate, in contrast to the smartphone, which is characterized as more engaging. In addition, condition C1 stood out as more accurate and robust, while condition C2 appeared to be easier to use.

Evaluation of distinct input methods of an intelligent wheelchair in simulated and real environments: a performance and usability study.

This article focuses on evaluating the usability of an intelligent wheelchair (IW) in both real and simulated environments. The wheelchair is controlled at a high-level by a flexible multimodal interface, using voice commands, facial expressions, head movements and joystick as its main inputs. A quasi-experimental design was applied including a deterministic sample with a questionnaire that enabled to apply the system usability scale. The subjects were divided in two independent samples: 46 individuals performing the experiment with an IW in a simulated environment (28 using different commands in a sequential way and 18 with the liberty to choose the command); 12 individuals performing the experiment with a real IW The main conclusion achieved by this study is that the usability of the IW in a real environment is higher than in the simulated environment. However, there were not statistical evidences to affirm that there are differences between the real and simulated wheelchairs in terms of safety and control. Also, most of users considered the multimodal way of driving the wheelchair very practical and satisfactory. Thus, it may be concluded that the multimodal interfaces enables very easy and safe control of the IW both in simulated and real environments.

Learning hybrid locomotion skills-Learn to exploit residual actions and modulate model-based gait control.

This work has developed a hybrid framework that combines machine learning and control approaches for legged robots to achieve new capabilities of balancing against external perturbations. The framework embeds a kernel which is a model-based, full parametric closed-loop and analytical controller as the gait pattern generator. On top of that, a neural network with symmetric partial data augmentation learns to automatically adjust the parameters for the gait kernel, and also generate compensatory actions for all joints, thus significantly augmenting the stability under unexpected perturbations. Seven Neural Network policies with different configurations were optimized to validate the effectiveness and the combined use of the modulation of the kernel parameters and the compensation for the arms and legs using residual actions. The results validated that modulating kernel parameters alongside the residual actions have improved the stability significantly. Furthermore, The performance of the proposed framework was evaluated across a set of challenging simulated scenarios, and demonstrated considerable improvements compared to the baseline in recovering from large external forces (up to 118%). Besides, regarding measurement noise and model inaccuracies, the robustness of the proposed framework has been assessed through simulations, which demonstrated the robustness in the presence of these uncertainties. Furthermore, the trained policies were validated across a set of unseen scenarios and showed the generalization to dynamic walking.