For a better quality of life
The global ageing population, along with disability compensation constitute major challenging societal and economic issues. In particular, achieving autonomy remains a fundamental need that contributes to the individual’s wellness and well-being, which could have a significant socio-economic impact. In this context, innovative and smart technologies are designed to achieve independence while matching user’s individual needs and desires.

For people suffering from motor disabilities (due to coordination limitations, dexterity incapacities, injury, accidents...), powered wheelchairs remain one of the most used assistive technologies, since it is synonym of freedom of navigation and travel. However, operating a wheelchair in a safe and reliable way requires cognitive skills (typically to anticipate obstacles and to plan a safe trajectory) as well as good visual-perceptual abilities. As a consequence, because of inadequate and dangerous reactions encountered while navigating, some disabled people are not allowed to drive electrical wheelchair, thus dramatically reducing their autonomy. Therefore, designing a robotic assistive solution related to wheelchair navigation remains of major importance as soon as it compensates partial incapacities.

In addition, in order to enhance the driving performances and the related user Quality of Experience, biofeedback can be necessary in the case where users suffer from visual and/or cognitive impairments and are not able to clearly observe their unsafe trajectory. It can be seen as a communication channel between the user and the wheelchair controller: such an active feedback can lead to minimal interferences from the automatic trajectory correction system as the user can gain awareness of the assistance provided by the system.

Then, coupling biofeedback solutions and driving assistance is of major importance in order to notify the user of danger and guide him over to a safer zone.

Research directions
This project focuses on two main complementary objectives:

• - the definition of an effective sensor-based indoor/outdoor wheelchair navigation framework: the idea is to secure the navigation by avoiding both static and dynamic obstacles, avoiding instabilities on sidewalks and preventing from falling off the curb.
• - the design of biofeedback solutions.

1. definition of a multimodal sensor platform able to detect vertical changes in level for safe navigation on sidewalk,
2. definition of an obstacle avoidance strategy in case of dynamic obstacles, including sensor-based servoing and tracking system,
3. definition of biofeedback platforms and interfaces adapted to neurological diseases
4. definition of a methodology related to clinical trials for evaluating both the system and the medical condition of the wheelchair user,
5. experiments within the PAMELA Lab in UCL and clinical trials within the Pôle Saint Hélier.

Challenges
In order to ensure a widespread use of the robotic system, defining a smart and low-cost driving assistance remains a challenging task, especially for outdoor navigation. Innovative sensors will have to be then properly designed in order to match robustness, large field of view and low-cost constraints.

In addition, innovative interfaces, enabling relevant biofeedback (medically validated), constitute a second major challenge. The correction of the trajectory, obtained thanks to the assistance module, will have to be perceived by the user by means of sensitive (visual, tactile…) feedback that will have to be easily adapted to the pathology.

Methodology
Naturally, the proposed assistance solutions have to be compliant with both the user needs and the medical advice. The rehabilitation center of Pôle Saint Hélier is specialized in the rehabilitation of disabled people suffering from neurological diseases and has designed rigorous methodologies related to clinical trials. From this, the idea is to determine relevant measures that can characterize the medical conditions of the patient that navigates with the proposed robotized wheelchair and the adapted biofeedback. These objective measures should help the medical staff to evaluate the evolution of a given pathology as well as the relevance of the proposed biofeedback solutions.

2016 - First year
All the research topics in the project are collectively studied by the partners, who will continue to advance beyond the state of the art. The objective of the project is to help the research teams associating their complementary approaches. During the first year, we will start working on tasks 1, 2, 3, 4 and 5 that are described in previous section.

Considering task 1, related to the definition of a multimodal sensor platform, we will focus our work on the study on innovative sensors for secure navigation on sidewalks. The definition of a set of (relatively) low-cost sensors able to detect changes in level or curbs remains a challenging task. Then, the proposed platform will be integrated onto the robotized wheelchair and a navigation framework on sidewalks will be designed.

In task 2, the navigation process will rely on a multimodal sensor based servoing. By coupling dynamic obstacles detection and tracking, with a more local obstacle avoidance framework, we intend to obtain a robust and reactive navigation assistance framework.

For task 3, during this first year, we will determine with the help of the medical staff the desired functionalities of an adapted interface and biofeedback. This will require the definition of a dedicated methodology of tests. This work will be realized in conjunction with task 4.

The results obtained in task 1 (navigation on sidewalks) will be reused and tested within the PAMELA lab platform (UCL) before realizing clinical trials with inpatients of the Pôle Saint Hélier.

2017 - Second year
Considering task 1, related to the definition of a multimodal sensor platform, we will continue the study of the foreseen solution for secure navigation on sidewalks. We plan to verify the ability of ultrasonic sensors platform to detect changes in level or curbs and to be coupled with a dedicated shared control law.

In task 2, we intend to augment the ability of the proposed shared control solutions by developing dynamic obstacles detection and tracking. As the solutions developed during the first year are complementary, we intend to blend them in order to propose a unified, robust and reactive navigation assistance framework.

For task 3, during this second year, thanks to the medical recommendations and the preliminary tests collected during first year, we will design an adapted haptic joystick as well as visual feedback. From this, we expect to enhance the user experience and to augment the cognitive abilities of users. To measure this, we have to design a methodology devoted to the evaluation of the impact of navigation assistance on the cognitive point of view (task 4) before realizing intensive clinical trials (task 5).

The results obtained in tasks 1, 2 and 3 will be reused and tested within the PAMELA lab platform (UCL) before realizing clinical trials with inpatients of the Pôle Saint Hélier. In addition, in order to improve the navigation on crowded sidewalks, tests will be conducted in order to determine how wheelchair user and a crowd can interact: from these tests, an interaction model will be determined and will be reused to enhance the navigation strategies.

2018 - Third year
Considering tasks 1 and 2, related to the safe navigation on sidewalks, we will continue our study of our proposed control law. We will first conduct trials with the medical staff of Pôle Saint Hélier in order to assess the quality of experience related to this driving assistance (tasks 4 and 5). We also intend to design a dedicated shared control law that able to safely cross curb ramps. To this aim, we first have to determine adequate features corresponding to the geometry of the ramp in order to define a sensor-based servoing process. ToF IR sensors and US sensors will be used. Experiments will be conducted both in PAMELA (UCL) and in the Pôle Saint Hélier with the collaboration of medical staff (tasks 4 and 5). In addition, from the analysis of the social navigation data recorded during tests performed both in Pamela and Ker Lann platform this year, we will design an additional navigation task for outdoor human-aware navigation.

One of UCL’s partner charities, MERU, is currently designing a new version of the Bugzi wheelchair for children, with the view of enabling it to work in outdoor environments. This will require significant improvements to its sensing and safety capabilities. Therefore, the ISI4NAVE team will test the suitability of our curb detection system on the Bugzi wheelchair platform. Finally, regarding tasks 3 and 5, we will continue the development and testing of shared control for alternative wheelchair interfaces, i.e. for people with medical conditions that prevent them from using a conventional joystick [3, 8]. This will include discrete inputs such as buttons and head arrays, sip-and-puff switches and Brain-Computer Interfaces, with tests taking place both at UCL and Rennes.