Section: New Results

Navigation of Mobile Robots

Visual Navigation from an Image Memory

Participants : Paolo Robuffo Giordano, François Chaumette.

This study achieved during Suman Raj Bista's Ph.D. was concerned with visual autonomous navigation in indoor environments. As in our previous works concerning navigation outdoors [4], the approach is based on a topological localization of the current image with respect to a set of keyframe images, but the visual features used for this localization as well as for the visual servoing are not composed of points of interest only, but on a combination of points of interest and straight lines since they are more common indoors [60]. Satisfactory experimental results have been obtained using the Pioneer mobile robot (see Section 6.8.2) and Pepper (See Section 6.8.4).

Robot-Human Interactions during Locomotion

Participant : Julien Pettré.

In collaboration with the Gepetto team of Laas in Toulouse and the Mimetic group in Rennes, we have studied how humans avoid collision with a robot. Understanding how humans achieve such avoidance is crucial to better anticipate humans' reactions to the presence of a robot and to control the robot to adapt its trajectory accordingly. It is generally assumed that humans avoid a robot just like they avoid another human. Last year, we brought the empirical evidence that humans actually set a specific strategy to avoid robots: they showed a preference to give way to a robot [36]. However, the robot was passive, i.e., not reacting to the presence of participants. This year, we studied interactions between humans and reactive robot, performing avoidance maneuvers to avoid collisions. Our conclusions are that, in such situations of human-robot interactions, human behave again as during human-human avoidance interactions. Again, this study provides useful guidelines about the design of robot control techniques.

Semi-Autonomous Control of a Wheelchair for Navigation Assistance

Participants : Louise Devigne, Marie Babel.

In order to improve the access to mobility for people with disabilities, we have previously designed a semi-autonomous assistive wheelchair system which progressively corrects the trajectory as the user manually drives the wheelchair and smoothly avoids obstacles. Within the frame of ISI4NAVE associated team (see Section 9.4.1.2), we investigated probabilistic blending approaches which take into account uncertainty in the interaction [45]. We also designed a shared-control curb-following solution for outdoor assisted power wheelchair navigation. Once a curb is detected, user input is blended with constraints deduced from the distance from sensors to the detected curb. This provides an intuitive shared control scheme capable of assisting the user while needed i.e. while approaching a curb. Preliminary validation tests of the robotic system were conducted within the PAMELA facility.

Developing and testing such systems for wheelchair driving assistance requires a significant amount of material resources and clinician time. With Virtual Reality technology, prototypes can be developed and tested in a risk-free and highly flexible Virtual Environment before equipping and testing a physical prototype. Additionally, users can "virtually" test and train more easily during the development process. We then designed a power wheelchair driving simulator allowing the user to navigate with a standard wheelchair in an immersive 3D Virtual Environment. In order to validate the framework including the driving assistance solution, we performed tests on the Immersia platform (Inria Hybrid team) with able-bodied participants and we have shown that the simulator it generates a good sense of presence and requires rather low cognitive effort from users [44].

Wheelchair Kinematics and Dynamics Modeling for Shared Control

Participants : Aline Baudry, Marie Babel.

The driving experience of an electric powered wheelchair can be disturbed by unpleasant dynamic effects of the caster wheels, particularly during maneuvers in narrow rooms and direction changes. In order to prevent their nasty behaviour, we propose to model caster wheel kinematics and dynamics in order to implement a control law for a semi-autonomous assistance to maneuver in narrow environments. We conducted a preliminary study that has been achieved for our three types of wheelchair, each presenting different kinematic behaviors: front caster type, rear caster type and mid-wheel drive (see Figure 3.c). Transfer functions for each of these configurations have been identified. We achieved to design a parametric transfer function of the caster's behavior regarding to the initial orientation, wheelchair's velocity and user mass, in order to develop a sensorless maneuver control law.

Wheelchair Autonomous Navigation for Fall Prevention

Participants : Solenne Fortun, Marie Babel.

The Prisme project (see Section 9.1.8) is devoted to fall prevention and detection of inpatients with disabilities. For wheelchair users, falls typically occur during transfer between the bed and the wheelchair and are mainly due to a bad positioning of the wheelchair. In this context, the Prisme project addresses both fall prevention and detection issues by means of a collaborative sensing framework. Ultrasonic sensors are embedded onto both a robotized wheelchair and a medical bed. The measured signals are used to detect fall and to automatically drive the wheelchair near the bed at an optimal position determined by occupational therapists. We first designed a detection solution based on a multiple echoes technique that enhances the system perception abilities. This augmented perception system is planned to be used for wheelchair navigation as well as fall detection.

Robotic Platform for Assistance to People with Reduce Mobility

Participants : Dayana Hassan, Paolo Salaris, Patrick Rives.

The main objective of this work is to develop, in collaboration with AXYN Robotics (see Section 8.2.4), an intelligent vehicle to help elderly or persons with reduced mobility to move safely within a retirement home, an hospital or other much more crowded and dynamic environments. First of all, the vehicle has to be able to move within the environment while at the same time update the current map as accurately as possible. Once the map of the environment is available, the robot has to be able to plan the trajectory and reach a given destination. The robot should also follow a person taking into account social behaviors or bring towards a given destination, e.g. the canteen, making sure that an elderly person, affected e.g. by Alzheimer's disease, follows the robot. The robot should also work as an intelligent walker and help people in case of falling. In all these cases, it is very important to include humans (i.e. his/her model, his/her behaviors, his/her intentions etc.) within the study in order to develop adaptable human-aware path planning and control strategies. During this first year, the problem of following a person has been studied, starting from the literature, in order to find a suitable control scheme that merges feedback control laws, aimed at reactively cope with neighborhood environment events and feedforward ones, mainly intended to take into account the intentions of the person to follow, also including social behaviors.